Model Database

From Upscale Wiki
This is the approved revision of this page, as well as being the most recent.
Jump to navigation Jump to search
If a model has no license, that means all rights reserved. You will need a private license or explicit permission from the creator to use it commercially or with modifications. However you may use all model here on non-commercial projects, that may not in any way cost money. If you gain permission to use a model in another way including commercially, you should still credit the author and link to the original model source.
For Datasets to train your models, go here: https://upscale.wiki/wiki/Dataset_Database
Looking for Official Models (Such as ESRGAN, BSRGAN, or Real-ESRGAN)? Check this page: https://upscale.wiki/wiki/Official_Research_Models

ESRGAN ("old Architecture") Models

These are all models that use the "old" ESRGAN architecture. There are various GUIs available to inference/upscale with these models. The only actively maintained program is chaiNNer by Joey. The others are: IEU by Honh Cupscale by NMKD.

These programs can be used to train your models: BasicSR, the official ESRGAN repository (old arch tag), victorca's traiNNer (https://github.com/victorca25/traiNNer), or sudo's colab-traiNNer (https://github.com/styler00dollar/Colab-traiNNer/).

Image scaling and Video upscaling

Universal Models

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
CountryRoads EzoGaming 4x WTFPL ESRGAN Universal Upscaler 2021-08-26 Streets with dense foliage in the background. Outdoor scenes. 4xPSNR https://cdn.discordapp.com/attachments/547949806761410560/880291337373618236/unknown.png, https://media.discordapp.net/attachments/547949806761410560/880290211026853918/unknown.png, https://media.discordapp.net/attachments/547949806761410560/880299665428447232/unknown.png, https://cdn.discordapp.com/attachments/880291340301254657/880303552315146340/unknown.png
Remacri Remacri Backup Mirror Foolhardy 4x CC BY-NC-SA 4.0 General Upscaler 2021-04-09 A creation of BSRGAN with more details and less smoothing, made by interpolating IRL models such as Siax, Superscale, Superscale Artisoft, Pixel Perfect, etc. This was, things like skin and other details don't become mushy and blurry. none - interpolated
4x-UltraSharp Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Universal Upscaler 2021-10-27 This is my best model yet! It generates lots and lots of detail and leaves a nice texture on images. It works on most images, whether compressed or not. It does work best on JPEG compression though, as that's mostly what it was trained on. It has the ability to restore highly compressed images as well! If you want a more balanced output, check out the UltraMix Collection down below. It's a bunch of interpolated models based around UltraSharp and my other models 4xESRGAN Sample 1 Sample 2 Sample 3 Sample 4
UniScale-Balanced/Strong Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Universal Upscaler 2021-08-23 UniScale strikes a nice balance between sharpness and realism. This model can upscale almost anything well. It was originally intended to upscale game textures, but was expanded into a universal upscaler. Interp is these two models interpolated. 4xESRGAN https://slow.pics/c/i75q0yN1
UniScaleNR-Balanced/Strong Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Universal Upscaler 2021-08-23 Version of UniScale trained with camera noise injection (NR = Noise Removal). This model removes noise from images while upscaling. 4xESRGAN https://slow.pics/c/UkCwdK13
UniScale_Restore Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Universal Upscaler 2021-08-27 UniScale_Restore has strong compression removal that helps with restoring heavily compressed or noisy images. It is intended to compete with BSRGAN. Trained with BSRGAN_Resize and Combo_Noise in traiNNer. 4xESRGAN https://cdn.discordapp.com/attachments/880826637543964712/880934732748169226/unknown.png, https://cdn.discordapp.com/attachments/880826637543964712/880932754102026290/unknown.png
UniScaleV2_Soft/Moderate/Sharp Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Universal Upscaler 2021-09-06 These models work great on game textures when interpolated 50/50 with UniScale_Restore, and work amazingly on uncompressed images. DO NOT USE FOR COMPRESSED IMAGES, use the original UniScale or UltraSharp for that. 4xESRGAN https://cdn.discordapp.com/attachments/884239326471393331/884296266857713704/unknown.png https://cdn.discordapp.com/attachments/884239326471393331/884294468709257216/unknown.png https://cdn.discordapp.com/attachments/884239326471393331/884298572894441512/unknown.png https://cdn.discordapp.com/attachments/884239326471393331/884302370136289323/unknown.png
realesrgan-x4minus DinJerr 4x WTFPL Real-ESRGAN General Upscaler 2022-04-19 Basically realesrgan-x4plus without the degradation training. Supposed to help retain more details, but unfortunately due to the dataset (I think) still blurs details adjacent to other objects. realesrgan-x4plus https://imgsli.com/MTA0OTA2
4x_2C2-ESRGAN_Nomos2K_200000_G.pth Joey 4x MIT 2C2-ESRGAN General Upscaler 2022-04-20 Technically my previous experiment was the pretrained model, but for all intents and purposes this was trained from scratch. Description: Pretrained model for the new architecture modification I made. You can read more about it in the github README. Basically it makes smaller ESRGAN models that theoretically can produce the same level of quality. NOTE: THIS WILL NOT WORK IN CUPSCALE, IEU, OR CHAINNER (yet)! You have to use my fork to use it for now.
Misc Alsa 4x CC BY-NC-SA 4.0 ESRGAN General Upscaler The Misc model is trained on various pictures shot by myself (Alsa), including bricks, stone, dirt, grass, plants, wood, bark, metal and a few others. Manga109Attempt
FuzzyBox BlueAmulet 4x CC BY-NC 4.0 ESRGAN General upscaler 2020-07-21 Photographs, Artwork, Textures, Anything really - Tried out a new pixel loss idea based on ensuring the HR downscaled matches the LR. Colors are pretty good as well as edges, but generated details seem slightly fuzzy hence the name. RRDB_ESRGAN_x4
Lollypop LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN General upscaler 2020-09-11 A universal model, that is aimed at prerendered images, but handles realistic faces, manga, pixel art and dedithering as well. Trained using the patchgan discriminator, with cx loss, cutmixup and frequency separation, it produces good results with a slight grain due to patchgan, with some sharpening using cutmixup. 4X_esrgan.pth
UniversalUpscalerV2 Mutin_Choler 4x WTFPL ESRGAN General Upscaler 2021-04-01 General Upscaler 1xESRGAN https://imgsli.com/NDc3ODA/2/1
NMKD Siax ("CX") Nmkd 4x WTFPL ESRGAN General upscaler 2020-11-06 Universal upscaler for clean and slightly compressed images (JPEG quality 75 or better) 4xPSNR

Realistic Photos

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
4x-Valar musl 4x CC0 ESRGAN+ Realistic Photos 2021-09-16 Meant as an experiment to test latest techniques implemented on traiNNer, including: AdaTarget, KernelGAN, UNet discriminator, nESRGAN+ arch, noise patches, camera noise, isotropic/anisotropic/sinc blur, frequency separation, contextual loss, mixup, clipL1 pixel loss, AdamP optimizer, etc. The config file is provided on the download link above. I encourage everybody to mirror the model, distribute and modify it in anywway you want. 4x_RRDB_ESRGAN Sample 1
4x_Box buildist 4x GNU GPLv3 ESRGAN Realistic Photos 2019-06-20 RRDB_ESRGAN_x4 replacement for stuff that's supposed to look realistic. none
Nickelback BlackScout 4x GNU GPLv3 ESRGAN Realistic Photos 2020-02-12 this model aims to improve further on what has been achieved by the regular 4xESRGAN and also 4xBox. It can upscale most pictures/photos (granted they are clean enough) without destroying as much detail as the aforementioned models. It generates less moiré like patterns and keeps details without oversharpening or blurring the image too much. 4xESRGAN
NickelbackFS BlackScout 4x GNU GPLv3 ESRGAN-FS Realistic Photos 2020-07-10 This model aims to improve further on what has been achieved by the old Nickelback which was an improvement attempt over 4xESRGAN and also 4xBox. It can upscale most pictures/photos (granted they are clean enough) without destroying as much detail as Box and basic ESRGAN. 4xESRGAN
4x NMKD Superscale Nmkd 4x WTFPL ESRGAN Clean Real-World Images 2020-07-22 Upscaling of realistic images/photos with noise and compression artifacts 4xESRGAN


Art/Pixel Art

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
4x_xbrz_90k LyonHrt 4x CC BY-NC 4.0 ESRGAN Pixel Art 2019-06-05 Xbrz style pixel art upscaler 4xPSNR
Rebout LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Pixel Art/Sprites 2019-08-04 For upscaling character sprites Detoon
Rebout Blend LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Pixel Art/Sprites Dataset: Custom prepared sprites from kof 94 rebout, which with gradients blending Detoon
Arzenal (v1.1) ComputerK 8x CC BY-NC-SA 4.0 ESRGAN Pixel Art Smooth general pixel art, Minecraft textures Interpolated from nmkd's pixel art upscaling models with different interpolation settings and de-dithered versions
4x_scalenx_90k LyonHrt 4x CC BY-NC 4.0 ESRGAN Pixel Art 2019-06-15 Scalenx style pixel art upscaler 4xPSNR
4x_xbrz+dd_260k LyonHrt 4x CC BY-NC 4.0 ESRGAN Pixel Art 2019-06-17 xbrz plus dedithering style pixel-art upscaling model to wiki under specialised 4x_xbrz_90k
Lady0101_208000.pth DinnJerr 4x ESRGAN Pixel Art/Paintings 2019-07-25 Upscale pixel art/paintings to digital painting style WaifuGAN_v3_30000.pth
Fatality (4x_Fatality_01.pth) twittman 4x ESRGAN Pixel Art/Sprites 2019-07-29 Upscales medium resolution Sprites, dithered or undithered, can also upscale manga/anime and gameboy camera images. 4x_Faces_04_N_180000_G.pth
Fatality MK2 (4x_Fatality_MKII_90000_G.pth) twittman 4x ESRGAN Pixel Art/Sprites Dataset: Anime, Manga and some real life A previously attempted MK2
Fatal Pixels (4x_FatalPixels_340000_G.pth) twittman 4x ESRGAN Pixel Art/Sprites Dataset: Anime, Manga Fatality_MKII_90k
Faithful 2x Joey 2x CC-BY-NC 4.0 ESRGAN Pixel Art 2019-09-08 2xESRGAN
deviantPixelHD raulsangonzalo 4x You're free to use this model for all non-commercial projects and interpolation, please give credit if featured in any projects ESRGAN Pixel Art 2019-09-09 Similar to Manga109, can be used as a general digital upscaler as well as with pixel art RRDB_PSNR_x4 https://www.youtube.com/channel/UCwBfuiHdSPQ-zslOmiW8OHg
4x_BS_Deviance systemd-resolved/BlackScout 4x GNU GPLv3 ESRGAN Art 2020-02-25 This model upscales Digital Drawings. It was trained on random drawings found on DeviantArt. Mostly Landscape and Scenery and Illustrations of Characters. It does fairly well and works on many different styles. 4xESRGAN
4xSmoothRealism Joey 4x CC BY-NC 4.0 ESRGAN Pixel Art 2020-03-04 Pixel art, rocky/grainy textures? Quantization smoothing, adding detail. 4x_RRDB_PSNR_old_arch.pth
8x_HugePeeps_v1 DinJerr 8x CC BY-NC 4.0 ESRGAN Art/People 2020-06-16 Painted humans TGHQFace8x
ArtStation DinJerr 8x ? ESRGAN Art/People 2019-08-21 GO UP A FOLDER, THEN TO THE ARTSTAION1000 OR ARTSTATION1337 FOLDER
4x_BS_DevianceV2 systemd-resolved/BlackScout 4x GNU GPLv3 ESRGAN-FS Art 2020-07-25 upscales Digital Drawings. It was trained on random drawings found on DeviantArt. Mostly Landscape and Scenery and Illustrations of Characters. It does fairly well and works on many different styles. 4xESRGAN
4x_BS_DevianceMIP systemd-resolved/BlackScout 4x GNU GPLv3 ESRGAN-FS Art 2020-07-25 upscales Digital Drawings. It was trained on random drawings found on DeviantArt. Mostly Landscape and Scenery and Illustrations of Characters. It does fairly well and works on many different styles. 4xESRGAN
4x_Struzan Laserschwert 4x CC BY-NC-SA 4.0 ESRGAN Art 2020-10-12 Upscaling airbrush/pencil-based artwork 4xPSNR https://drive.google.com/drive/folders/1fyeIWInDrM6r-xxrCW4U09oafWw08u9S?usp=sharing
4x_PixelPerfectV4_137000_G Mutin Choler 4x WTFPL ESRGAN Pixel Art/Sprites 2020-11-16 Sprite Upscaler 4xESRGAN https://imgsli.com/MzgxMTc/1/2
BigFArt DinJerr 4x CC BY-NC 4.0 ESRGAN Pixel art 2020-15-31 Larger-scaled pixels to digital painting Face Focus
MS Unpainter FoolhardyVEVO 8x CC BY-NC-SA 4.0 ESRGAN Pixel Art 2021-03-24 Low-resolution MS Paint drawings, general pixel art, and general dithered pixel art, all kinds of pixel art A mix of: 8x_glasshopper_ArzenalV1.1, 8x_glasshopper_MS-Unpainter, 8x_NMKD-Sphax + Sphax de-dither, and 8x_NMKD-YanderePixelArt4 + Yandere De-Dither, all interpolated with adjustments using the interpolate function in Cupscale
MS Unpainter De-Dither FoolhardyVEVO 8x CC BY-NC-SA 4.0 ESRGAN Pixel Art 2021-03-24 Low-resolution MS Paint drawings, general pixel art, and general dithered pixel art, all kinds of pixel art A mix of: 8x_glasshopper_ArzenalV1.1, 8x_glasshopper_MS-Unpainter, 8x_NMKD-Sphax + Sphax de-dither, and 8x_NMKD-YanderePixelArt4 + Yandere De-Dither, all interpolated with adjustments using the interpolate function in Cupscale
UltraFArt_v3 / FArtDIV3 Suite DinJerr 4x CC BY-NC-SA 4.0 ESRGAN Art 2021-05-14 Painting style with larger shaped features (?). 4x_UltraFArt https://1drv.ms/u/s!Aip-EMByJHY27UIFr1mk87zFytap?e=ZZliII
NXbrz Archerpolation 4x CC BY-NC-SA 4.0 ESRGAN Pixel Art 2021-06-09 Basic pixel art upscaling, for people who want a more simpler style and lightweight pixel art upscaling model. https://drive.google.com/drive/folders/1mYTMpwDlKQulBmjgKpcxL3-T6xAGXVxl?usp=sharing

Drawn Material

Anime

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


4x-AnimeSharp Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Anime or Text 2021-12-28 Text or Anime Interpolation between 4x-UltraSharp and 4x-TextSharp-v0.5. Works amazingly on anime. It also upscales text, but it's far better with anime content. I rebranded this model on 2/10/22 to 4x-AnimeSharp from 4x-TextSharpV1. Text Sample Anime Sample Anime Sample 2
4x-AnimeSharp-lite Kim2091 4x CC BY-NC-SA 4.0 ESRGAN-lite Anime 2022-02-11 Anime This model is a lite version of AnimeSharp. It was trained using student-teacher learning (if i'm using the term properly), where the HRs are LRs upscaled by the full size AnimeSharp ESRGAN model, and the lite model is trained on those outputs as the HR. It works best on clean or slightly blurry anime. Downscale by 50% first in almost all cases https://cdn.discordapp.com/attachments/547949405949657100/941776512049360986/unknown.png https://cdn.discordapp.com/attachments/547949405949657100/941770824543797268/unknown.png https://cdn.discordapp.com/attachments/903415274521374750/941779970013925506/unknown.png
sudo_RealESRGAN2x_3.332.758_G.pth / sudo_RealESRGAN2x_Dropout_3.799.042_G.pth (pcloud) sudo_RealESRGAN2x_3.332.758_G.pth / sudo_RealESRGAN2x_Dropout_3.799.042_G.pth (mediafire) sudo 2x CC BY-NC-SA 4.0 ESRGAN (6B) Anime 2022-06-25 Tried to make the best 2x model there is for drawings. I think i archived that. And yes, it is nearly 3.8 million iterations (probably a record nobody will beat here), took me nearly half a year to train. It can happen that in one edge is a noisy pattern in edges. You can use padding/crop for that. I aimed for perceptual quality without zooming in like 400%. Since RealESRGAN is 4x, I downscaled these images with bicubic. I would recommend my VSGAN code though and just load the onnx. https://github.com/styler00dollar/VSGAN-tensorrt-docker I just wanted a good 2x model for animations, but that model can also be used for wallpapers and so on. Before I hear people complaining, the dropout model is a modified architecture. Stuff like cupscale or chaiNNer won't work with pth. Load the onnx with VSGAN or chaiNNer. I did add the model before switching to dropout though, which is normal ESRGAN pth, that one should work everywhere. I also converted everything into onnx, jit and ncnn, so pretty much everything there is. If you want to use ncnn, don't use nihuis code (that also includes cupscale), these codes don't include propper tiling in C++, which is very bad for this model. I think chaiNNer should have overlap/padding with ncnn, so use that instead if you really want ncnn. Plz don't steal without credits, k thx. Pretrained_Model_G: RealESRGAN_x4plus_anime_6B.pth / RealESRGAN_x4plus_anime_6B.pth (sudo_RealESRGAN2x_3.332.758_G.pth) https://cdn.discordapp.com/attachments/579685650824036387/990323693660033045/jpg_compare.png https://cdn.discordapp.com/attachments/579685650824036387/990323694196916275/test2_compare.png https://cdn.discordapp.com/attachments/579685650824036387/990323694561816606/f4_compare.png https://cdn.discordapp.com/attachments/579685650824036387/990323695186763884/test_compare.png https://cdn.discordapp.com/attachments/579685650824036387/990323695601987584/f2_compare.png
2x_Bubble_AnimeScale_Compact_v1 Bubblemint#6472 2x CC BY-NC-SA 4.0 Real-ESRGAN Compact Anime or Text 2022-11-13 Text or Anime This is my first model, so it's not perfect, but I wanted to see if I could train an upscaling model that didn't result in a lot of detail loss and deblurring like the current Compact upscaling models. I believe I accomplished this, but I was unable to reduce contrast shifting. The contrast shifting may cause skin tones to appear incorrect on bright frames, but it's not too bad overall! I'll list a few examples below; more can be found by clicking the Overview link on the Github release page. https://imgsli.com/MTM0MzMx https://imgsli.com/MTM0MzM3 https://imgsli.com/MTM0MzQw
Sol Levante NTSC2HD Phoenix 4x UNLICENSE ESRGAN Anime/Pretrained 2020-04-08 NTSC DVD-spec encode x4 scale super-resolution for Anime Drawing style content. The dataset has a LOT of data throughout almost every frame, so it had a lot of stuff to learn. The resulting DVD-spec encode also had some blocking at times so it also learned to fight off blocking. RRDB_PSNR_x4
2x_AnimeClassics_UltraLite_510K CG1989 2x CC BY-NC-SA 4.0 ESRGAN Anime/Pretrained 2022-03-08 A 2x Ultra Lite model coming in under 8MB. Trained with over 15 sets of LRs ranging in a wide amount of issues. Handles Rainbows, Dot Crawl, MPEG/H.264 Compression, and may even assist in removing halos, and fixing blurriness in certain cases. This is my first public model for everyone. Best when used on old anime that is grainy. I can't say what anime it's best suited for as I have tried multiple series, and have found it does a good job on most all the tests. I wouldn't say use this for Western Animation, but it may work. I have done a few tests that I have shown in the upscale results, but that was chained with other models to achieve such a result. This model is meant to retain the more natural look of a series. There is a color shift on the end result, not drastic, but still noticable. I figure you should fix any color issues in post that way to give a more polished upscale. Big thanks to @SaurusX for the model name, and just helping out in general with anything. 2X_DigitalFilmV5_Lite.pth Sample 1 Sample 2 Sample 3 Sample 4
600k 650k BooruGan Tal 4x WTFPL ESRGAN Anime 2021-11-26 This model is designed to mainly upscale anime artworks. If you have issues with chroma then try the 600k iterations release. 4x_Manga109Attempt Sample 1 Sample 2 Sample 3
2x_Byousoku_5_Centimeter.pth Mystery_Bullet#0642 (Shady Adel) 2x GNU GPLv3 ESRGAN Anime/Pretrained 2020-04-08 Anime landscape upscale. Trained on Frames from BluRay of Byousoku 5 Centimeter 2xESRGAN
4x_OLDIES_290000_G_FINAL solidd93110 4x WTFPL ESRGAN Anime 2020-08-16 i made this model to upscale old anime and denoise. RRDB_PSNR_x4.pth
fidelbd_pokemodel Neo-Raws#4055 2x WTFPL ESRGAN Anime 2020-08-29 Made this model to upscale old anime that looks blurry. 2xESRGAN https://imgsli.com/MjExMjQ/
4x_OLDIES_ALTERNATIVE_FINAL.pth solidd93110 4x CC BY-NC-SA 4.0 ESRGAN Anime 2020-09-01 This model was made for my project captain tsubasa anime so i don't know if it works good for anything else. just try it ;) RRDB_PSNR_x4.pth https://imgsli.com/MjEzMDU
2x_SHARP_ANIME_V1 solidd93110 2x CC BY-NC-SA 4.0 ESRGAN Anime 2020-09-12 this model has been trained to work on lines and details - works well on animes which have fairly fine lines at the base but also the video must be progressive or deinterlaced 2xPSNR.pth https://imgsli.com/MjIwMDY
MeguUp /katoumegumi_#3231 GNU GPLv3. ESRGAN Anime 2020-09-16 Upscaling of lossless (uncompressed) anime art. LyonHrt (bloc97 dataset, MIT license). Fatal Pixels, Fatality Anime: Twittman (CC BY-NC-SA 4.0, acquired special permission to relicense. https://s.katou.pw/Discord_dPUiI1x3Pf.png). Deviance: |BlackScout (GNU GPLv3)
NMKD UltraYandere Nmkd 4x WTFPL ESRGAN Art/Anime 2020-10-08 Highly flexible 2D Art upscaling 4xESRGAN
NMKD UltraYandere Lite Nmkd 4x WTFPL ESRGAN Lite [nf=32 nb=12] Anime 2020-10-15 Fast Anime/Art upscaling 4x_DIV2K-Lite
2x_pokemodel_lite_100000_G Neo-Raws#4055 2x WTFPL ESRGAN Anime 2020-11-01 Upscale old anime like pokemon none https://imgsli.com/Mjc3Nzk https://imgsli.com/Mjc3Nzg https://imgsli.com/Mjc3ODA https://imgsli.com/Mjc3ODE https://imgsli.com/Mjc3ODI
2x_SHARP_ANIME_V2 solidd93110 2x CC BY-NC-SA 4.0 ESRGAN Anime 2020-12-09 2x_PSNR.pth https://imgsli.com/MzI0MDk
2x_BIGOLDIES_415000_G.pth solidd93110 2x CC BY-NC-SA 4.0 ESRGAN Anime 2020-12-09 upscaling old anime. help to denoise and find lines and dehalo 2x_PSNR.pth https://imgsli.com/MzI0MDc
NMKD YandereNeo Nmkd 2x WTFPL ESRGAN (Lite) Anime 2021-01-26 https://i.imgur.com/oxs71v5.png


NMKD YandereNeo Nmkd 4x WTFPL ESRGAN (Lite) Anime 2021-01-26 4x_DIV2K-Lite_1M https://i.imgur.com/oxs71v5.png
2x_Waifaux-NL3-SuperLite Joey 2x CC BY-NC-SA 4.0 ESRGAN 2021-02-23 Trained this model to see how it would work trying to essentially get the same results as Waifu2x but with ESRGAN none
2x_Waifaux-NL3-SRResNet Joey 2x CC BY-NC-SA 4.0 EDSR (SRResNet) 2021-02-25 Emulating Waifu2x at Noise Level 3 NOTE: You can't use this with regular esrgan forks or the bot, it has to be run through basicsr none
4x_Training4Melozard_Anime Joey 4x Whatever @B.Melozard2 wants ESRGAN Anime 2021-03-17 RRDB_ESRGAN_x4_old_arch
2x_LD-Anime_Skr_v1.0 Skr 2x CC BY-NC-SA 4.0 ESRGAN Denoise/Dehalo 2021-04-17 Denoise, dehalo, derainbow old anime 2xESRGAN https://imgsli.com/NTA3MDU/
2x_KemonoScale_v2 EzoGaming 2x CC BY-NC-SA 4.0 ESRGAN Anime 2021-06-05 Upscaling frames from Irodori anime (namely kemono friends) from 540p (the source render resolution) to 1080p, low resolution flat shaded art, de-JPEG of the aforementioned x2_CGIMaster_v1 https://cdn.discordapp.com/attachments/705815801290293329/850614850735439892/unknown.png
4x_muy4_035_1.pth katoumegumi_ 4x WTFPL ESRGAN Anime 2021-07-18 Upscaling of anime art (Specifically visual novel CG art) https://cdn.knightlab.com/libs/juxtapose/latest/embed/index.html?uid=278999ec-e802-11eb-abb7-b9a7ff2ee17c (Hakurei Reimu from Touhou Project illustrated by "okawa friend") https://cdn.knightlab.com/libs/juxtapose/latest/embed/index.html?uid=f3d4781a-e801-11eb-abb7-b9a7ff2ee17c (Kitaooji Karen from Making * Lovers developed by SMEE.)(edited)
Falcon Fanart LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Anime not on discord 4xPSNR

Manga

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
FSMangaV2 Jacob_ 4x CC BY-NC-SA 4.0 ESRGAN Manga 2020-12-14 Manga-style images with or without dithering - cartoons, maybe pixel art, etc RRDB_PSNR_x4 https://i.imgur.com/oVBVthF.png
4x_eula_digimanga_bw_v1_860k end user license agreement#9756 4x CC BY-NC-SA 4.0 ESRGAN Manga 2022-1-26 Black and white digital manga with halftones. 4xPSNR Clean Heavy Degradation
4x_eula_digimanga_bw_v2_nc1_307k end user license agreement#9756 4x CC BY-NC-SA 4.0 ESRGAN Manga 2022-08-17 Vast improvement over v1 in low frequency detail; moiré and artifacting reduced significantly and less random noise from JPEG artefacts in the input. Also now only works on 1 channel images, so it runs slightly faster on average and resulting images are much smaller but it might not work on some ESRGAN implementations, I personally recommend using chaiNNer. v1 may still be better in some edge cases.

There's also a supplementary 1x model that denoises very low quality LRs and smooths halftones so the image works better with the 4x model. Only trained it to help build the dataset and it's useless for already decent-ish LRs but may help you in some situations.

4x_eula_digimanga_bw_v1_860k https://slow.pics/c/Asbu0xgz
MangaScaleV3 Bunzero++ 2x CC BY-NC-SA 4.0 ESRGAN+ Manga 2021-05-24 To upscale manga including halftones, instead of trying to smooth them out. unreleased model https://imgsli.com/NTUyMjg

Cartoons

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_BroadcastToStudioLite_485k SaurusX 1x CC BY-NC-SA 4.0 ESRGAN Animation 2022-03-05 Improvement of low-quality cartoons from broadcast sources. Will greatly increase the visual quality of bad broadcast tape sources of '80s and '90s cartoons (e.g. Garfield and Friends, Heathcliff, DuckTales, etc). Directly addresses chroma blur, dot crawl, and rainbowing. You're highly advised to take care of haloing beforehand in your favorite video editor as the model will not fix it and may make existing halos more noticeable. Sample 1 Sample 2 Sample 3
2x_SwatKats_154000_G SaurusX 2x ESRGAN Animation 2021-11-17 In addition to removing the vertical blur, the model upscales, sharpens and will remove MPEG-2 artifacting and a small amount of rainbowing and dot crawl. Another series afflicted with the vertical blur is Avatar the Last Airbender, which can be repaired by this model. The video fed into the model MUST be 540 vertical for the deblur to work properly. 2xESRGAN https://imgsli.com/NzM2MDY https://imgsli.com/ODIxMjU
1x_SwatKatsLite_360000_G.pth SaurusX 1x ESRGAN Animation 2021-11-17 Fix vertical blur / split lines / shadowing. A 1x lite model of my 2xSwatKats. Resolves the same video problems as before, but 1x and faster and meant for chaining to other 2x models (or whatever). Input MUST be 540 vertical as the blur problem is very resolution sensitive. 2xESRGAN https://slow.pics/c/F4ilU8PM
1x_SheeepIt! Kim2091 1x CC BY-NC-SA 4.0 ESRGAN Animation 2022-07-21 Restoring frames from the show "Sheeep" This model was trained to restore "Sheeep" while retaining and enhancing the noise present in the show. The model amazingly finished training in only 8.8k iterations with no pretrain, with a dataset of only 4 image pairs. This model should work well for animes and cartoons with a lot of grain present. There are some slight haloing issues in dark colors unfortunately, but I was unable to fix it. Sample 1

Sample 2 Sample 3 Sample 4

2x-UniScale_CartoonRestore-lite Kim2091 2x CC BY-NC-SA 4.0 ESRGAN Animation, Pixel Art 2021-08-31 This model has VERY strong compression removal and line restructuring that allows it to restore any heavily compressed drawings, animation, cartoons, or anime. Also works on games as well as DDS compression. It renders frames very quickly and is very viable for restoring videos. 4x-MMScale Comparisons:

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Video Sample Original Video

2x_ATLA_KORRA_336200_G.pth aptitude 2x ESRGAN Animation 2021-07-22 Upscaling of Animation based on The Legend of Korra. 2xESRGAN.pth https://imgsli.com/NjIzMzU https://imgsli.com/NjIzMTY https://imgsli.com/NjIzMTc https://imgsli.com/NjIzMTg https://imgsli.com/NjIzMTk https://imgsli.com/NjIzMjA https://imgsli.com/NjIzMjE https://imgsli.com/NjIzMjQ https://imgsli.com/NjIzMjc https://imgsli.com/NjIzMjY https://imgsli.com/NjIzMjU
8x_BoyMeBob-Redux Joey 8x CC BY-NC-SA 4.0 ESRGAN+ (Joey's fork, eFonte fork, or iNNfer required to use) Animation 2021-08-19 Upscaling cartoons 8xBoyMeBob (unreleased) which used 8xESRGAN Examples
SpongeBob (4x_SpongeBob_235000_G / 4xSpongebob) Joey 4x CC-BY-NC 4.0 ESRGAN Animation 2019-09-12 4xESRGAN
Spongebob De-Quantize Joey 1x CC-BY-NC 4.0 ESRGAN Animation - Quantized 2019-11-02 Removed color quantization/indexing and dithering from cartoon style images and textures 1x_1xDEDITHER_32_512_126900_G
4x Spongebob v6 Deblur Joey 4x CC-BY-NC 4.0 ESRGAN Animation 2019-11-08 After not being entirely happy with the main Spongebob v6 model, I trained a new one with blurring OTF options and two different downscale types. This one is much better in my opinion. 4x Spongebob v6
4x Spongebob v6 De-Quantize Joey 4x CC-BY-NC 4.0 ESRGAN Animation - Quantized 2019-11-08 A model I trained to do both de-quantizing as well as upscaling. The results are pretty blurry but it works decently for what it is. 4x Spongebob v6
4x Spongebob v6 Joey 4x CC-BY-NC 4.0 ESRGAN Animation 2019-11-08 New version of the Spongebob model. Ideally it's a lot sharper and cleaner but I'm still not sure if it works better than the old one. From what I can tell it's better in many cases. 4xESRGAN
SpongeBob.CEL.2.HD.125ki.499e-PHOENiX.pth Phoenix 4x GPLv3 - SHARE YOUR CHANGES - If you use this as a pretrained model, share your new model!!! ESRGAN Animation 2019-12-24 Restorative CEL Animation MPEG-1 and 2 Model specifically crafted for SpongeBob S01. RRDB_PSNR_x4.pth
1x_SpongeBC1-Lite Joey 1x CC BY-NC-SA 4.0 BC1/DXT1 compression 2020-11-01 First ever lite BC1/DXT1 model. Probably only useful for cartoony textures like those in spongebob games or other cartoon licensed games. 1x_DIV2K-Lite_80k
4x_SpongeBob-Reloaded Joey 4x CC BY-NC-SA 4.0 ESRGAN Animation, Pixel Art 2020-12-11 4xSpongeBob
4x_SpongeBob-Reloaded-SWAG Joey 4x CC BY-NC-SA 4.0 ESRGAN Animation, Pixel Art 2020-12-11 4xSpongeBob

Digital Animation

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
4x_DigitalFake-2.1 Joey 4x CC BY-NC-SA 4.0 ESRGAN Digital Animation 2021-02-18 Replica of DigitalFrames 2.1 but interpolatable RRDB_ESRGAN_x4_old_arch.pth
4X_KCJPUNK_1.0233089 G.pth KCJPUNK 4x CC BY-NC-SA 4.0 ESRGAN Digital Animation 2021-03-23 Up-scaling Digital Animation 4x_fatal_Anime_500000_G
2X_KcjpunkAnime_2.0_Lite_196496_G KCJPUNK 2x CC BY-NC-SA 4.0 ESRGAN-lite Digital Animation 2021-03-23 Up-scaling Digital Animation This is my first attempt to make Light model so I started with 2x version. This model is much faster and give better results than my previous one.
2x_DigitoonLite_216k SaurusX 2x CC BY-NC-SA 4.0 ESRGAN-lite Digital Animation 2022-07-05 Meant as a versatile model for upscaling high detail digital anime and cartoons. Has debanding, MPEG-2 correction, and halo reduction. Trained to handle both 4:3 and 16:9 DVD material with equal efficacy. Will retain a lot of textures except for the really high freq stuff. 2x_DigiGradients_Lite_486k.pth https://imgsli.com/MTE1Mzg4
2x_DigiGradients_Lite_486k SaurusX 2x CC BY-NC-SA 4.0 ESRGAN-lite Digital Animation 2022-08-24 A very focused model meant for upscaling the TMNT 2003 DVDs. Degradations were added via AVISynth in order to match the video on the TMNT 2003 DVDs to correct the source problems. Problems corrected include aliased red chroma, chroma vertical blur, bad deinterlacing, banding, compression "grain", and poor animation line detail. The AVS scripts for the LR's were run through HCEnc to get authentic low bit rate MPEG-2 artifacts for fixing. By design, the final model gives a very digital-looking result and does not do a good job of retaining textures as the style of TMNT 2003 is all flats and gradients. none https://imgsli.com/MTAxNjc1 https://imgsli.com/ODQ2NzI

Drawings

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


KDM003_scans_1x.pth BoxDrop 1x ESRGAN Art 2019-07-29 Clean up model for scanned illustrations. - Made to remove moire patterns, reduce small imperfections, and correct mild compression artifacts in scanned Kingdom Death: Monster illustrations. CMYK printing often shifts colors, so this is intended to reverse that color shifting as well. Failed attempts based on ESRGAN_1x_JPEG_80to100
DigiPaint Rastrum 4x CC0 ESRGAN Art 2019-09-07 Digital Art Upscaler 4xfalcoon300(manga).pth

General Animation

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
American Dad 2 (American.Dad.2.HD.150ki.5e-PHOENiX.pth) PHEONiX (?) 4x ESRGAN Animation American Dad
FatalimiX Twittman 4x CC BY-NC-SA 4.0 (?) ESRGAN Animation 2019-10-29 Comic and Cartoon style images 4x_Fatality_MKII_90000_G_02.pth
Comic Book LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Animation Comic / Drawings. Trained on Custom (Spider-Man) dataset none (no interpolation)
Fatal_Anime Twittman 4x CC BY-NC-SA 4.0 (?) ESRGAN Animation Trained on Anime and Manga images none (no interpolation)
HugePaint DinJerr 8x CC BY-NC 4.0 ESRGAN Digital Illustrations Trained on a variety of images from ArtStation HugePeeps
4x_AbeScale mitch#1337 4x GNU ESRGAN Linework Cartoons Dataset: LR: Clone High DVD, HR: Re-illustrated Vectors from Frames 4xESRGAN
2x_Loyaldk-SuperPony_370000 V1.0 ChaseMMD 2x CC BY-NC-SA 4.0 ESRGAN Animation 2021-02-24 Upscale MLP episodes 2x_ERGAN https://imgsli.com/NDIwNTY https://imgsli.com/NDIwNTg https://imgsli.com/NDIwNjE https://imgsli.com/NDIwNjI https://imgsli.com/NDIwNjM
2x_Loyaldk-LitePony_380500 V1.0 ChaseMMD 2x CC BY-NC-SA 4.0 ESRGAN Animation 2021-03-01 Upscale MLP episodes. Liter version. 2x_ESRGAN
2x_Loyaldk-LitePony_500000_V2.0 ChaseMMD#6957 2x CC BY-NC-SA 4.0 ESRGAN (NF=32, NB=12) Animation 2021-04-06 Upscale MLP episodes. Liter version. Able to handle compression better compared to V1.0 and no longer creates halo's/rainbows. Good for Vector 2D art however converts detail to blobs. none
2x_Loyaldk-MediumPony_500000_V2.0 ChaseMMD 2x ESRGAN (NF=48, NB=18) Animation 2021-04-11 Upscale MLP episodes. Liter version. Able to handle compression better compared to V1.0 and no longer creates halo's/rainbows. Good for Vector 2D art however converts detail to blobs. Leaves more blobs when converting detail to blobs compared to litepony none
4x_Loyaldk-LitePony_500000_V2.0 ChaseMMD 4x CC BY-NC-SA 4.0 ESRGAN (NF=32, NB=12) Animation 2021-04-12 Good for Vector 2D art however converts detail to blobs. Quality seems weak with this one. However posting anyways to see if a use is found. none
2x_Loyaldk-SuperPony_500000_V2.0 ChaseMMD 2x CC BY-NC-SA 4.0 ESRGAN Animation 2021-04-30 Upscale MLP episodes. Full version. Able to handle compression better compared to V1.0 and no longer creates halo's/rainbows. Good for Vector 2D art however converts detail to blobs. none
4x_Loyaldk-SuperPony_500000_V2.0 ChaseMMD 4x CC BY-NC-SA 4.0 ESRGAN Animation 2021-04-30 Upscale MLP episodes. Full version. Able to handle compression better compared to V1.0 and no longer creates halo's/rainbows. Good for Vector 2D art however converts detail to blobs. none
4x_Loyaldk-MediumPony_500000_V2.0 ChaseMMD 4x CC BY-NC-SA 4.0 ESRGAN Animation 2021-04-30 Upscale MLP episodes. Full version. Able to handle compression better compared to V1.0 and no longer creates halo's/rainbows. Good for Vector 2D art however converts detail to blobs. none
Platoon Model Set ChaseMMD 4x CC BY-NC-SA 4.0 ESRGAN Animation 2021-06-07 Upscale Anime while keeping as much of the original detail as possible. Isn't among the sharpest models and if the anime is somewhat blurry it will retain that detail but add more pixels to give a smoother edge. Doesn't change the color very much and may make some scene's very slightly darker. Pony Models https://mega.nz/folder/5UAlQCRJ#2Bj1lq8PYeB3P9k5UKSvkQ


Traditional Animation

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
DigitalFilmV5 Lite OptimusPrimal 2x WTFPL ESRGAN Traditional Animation 2021-02-19 Upscaling Dragon Ball Z DBox DVD's. - grainy sources. It keeps some grain, but also doing some cleaning, sharpening, and fixing. none https://imgsli.com/NDE0MzQ https://imgsli.com/NDE0MzU https://imgsli.com/NDE0MzY https://imgsli.com/NDE0Mzc https://imgsli.com/NDE0Mzg https://imgsli.com/NDE0Mzk
4x_HellinaCel VGDCKeroro 4x fuckin do whatever ESRGAN Traditional Animation 2021-03-21 A rougher alternative to 4xCelFrames with a focus on realistic looking cels over nice looking cels. It's trained on DetoriationFrames LRs, so if you give it an image straight from the source it will go twice as hard on it. This can be used to your advantage, though. I recommend cleaning it up before running DetoriationFrames, or else it will come out rough. 4xESRGAN https://imgsli.com/NDU2MjU


Image Restoration

Image Compression

JPEG Artifacts
Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
JPG (00-20%) BlueAmulet 1x CC BY-NC 4.0 ESRGAN JPEG Trained on Custom (CC0 Textures) JPEG models by Alsa
JPG (20-40%) BlueAmulet 1x CC BY-NC 4.0 ESRGAN JPEG Trained on Custom (CC0 Textures) JPEG models by Alsa
JPG (40-60%) BlueAmulet 1x CC BY-NC 4.0 ESRGAN JPEG Trained on Custom (CC0 Textures) JPEG models by Alsa
JPG (60-80%) BlueAmulet 1x CC BY-NC 4.0 ESRGAN JPEG Trained on Custom (CC0 Textures) JPEG models by Alsa
JPG (80-100%) BlueAmulet 1x CC BY-NC 4.0 ESRGAN JPEG Trained on Custom (CC0 Textures) JPEG models by Alsa
JPG (00-20%) Alsa 1x CC BY-NC-SA 4.0 ESRGAN JPEG Trained on Custom (Photos / Manga)
JPG (20-40%) Alsa 1x CC BY-NC-SA 4.0 ESRGAN JPEG Trained on Custom (Photos / Manga)
JPG (40-60%) Alsa 1x CC BY-NC-SA 4.0 ESRGAN JPEG Trained on Custom (Photos / Manga)
JPG (60-80%) Alsa 1x CC BY-NC-SA 4.0 ESRGAN JPEG Trained on Custom (Photos / Manga)
JPG (80-100%) Alsa 1x CC BY-NC-SA 4.0 ESRGAN JPEG Trained on Custom (Photos / Manga)
Kim2091_DeJpeg_v0 Kim2091 1x CC BY-NC-SA 4.0 ESRGAN Superlite JPEG Model I forgot to release. This doesn't totally remove JPEG artifacts, but it does a decent job at a fast rate. It seemingly does a better job of retaining detail than some other JPEG models. The model is incomplete, I need to train it further on compact rather than ESRGAN. This is just a temporary release Custom JPEG dataset https://cdn.discordapp.com/attachments/903415274521374750/1041539960672632832/1668392853.8196685.png https://cdn.discordapp.com/attachments/903415274521374750/1041540021175463956/1668392885.0326388.png
4x NMKD-PatchySharp nmkd 4x WTFPL ESRGAN (Old Arch) Art/CGI 2020-09-12 Upscaler for clean images or images with compression artifacts (jpeg quality >75) - Produces very sharp lines/edges due to NN-Filtered HR images. Proven to produce very, very good results on drawings (sharp lines) and CGI, but should also work pretty well for real-world images. 4xESRGAN https://i.imgur.com/PzFkmAg.png
1x_JPEG (by compression level) BlueAmulet 1x CC BY-NC 4.0 ESRGAN JPEG 2019-08-01 A collection of models for jpeg artifact removal, examples are included in the provided link Alsa https://drive.google.com/open?id=1HWokMsUwsR_Mw-NOJgc8OdS8mkdIje_C
DeJpeg Fatality_PlusULTRA! Twittman 1x CC BY-NC-SA 4.0 (?) ESRGAN JPEG 2019-10-29 1x_DeJpeg_Fatality_01_200000_G.pth
DeCompress Kim2091 4x CC BY-NC-SA 4.0 ESRGAN JPEG (works on many compression formats) 2021-09-07 USE 4x-UltraSharp INSTEAD - This is UniScaleV2, but intended for images with compression. Seems to work best on realistic images. 4xESRGAN https://cdn.discordapp.com/attachments/884864597310459996/884952891721383936/unknown.png https://cdn.discordapp.com/attachments/884864597310459996/884957054807179274/unknown.png https://cdn.discordapp.com/attachments/884864597310459996/884960697883193354/unknown.png https://cdn.discordapp.com/attachments/884864597310459996/884953810097827850/unknown.png
JPG PlusULTRA (1x_jpg_PlusULTRA_130000.pth) Twittman 1x CC BY-NC-SA 4.0 (?) ESRGAN JPEG 2019-10-29 1x_DeJpeg_Fatality_01_200000_G.pth
SaiyaJin DeJpeg Twittman 1x CC BY-NC-SA 4.0 (?) ESRGAN JPEG 2019-10-29 Dataset: Anime, real life, manga, generated 1x_DeJpeg_Fatality_PlusULTRA_200000_G.pth
1x_JPEGDestroyer BlackScout 1x GNU GPLv3 ESRGAN JPEG 2020-02-19 This model is meant to reduce or eliminate JPEG Compression without making the original images too smooth or killing detail. It manages to do a fairly good job but don't expect overly compressed images to work with this. 1xESRGAN + previous attempts
4xJaypeg90 Jacob 4x GNU GPLv3 ESRGAN JPEG compression 2020-07-20 Photos/realistic 3D with JPEG compression, quality 85-95 and 4:2:0 chroma subsampling. - Created for Myst3 images, since all have 4:2:0 chroma subsampling and existing JPEG models did not give good results. Favors smoothing over over-sharpening. 4xESRGAN https://i.imgur.com/n09m2Rt.png
1x_SBDV-DeJPEG-Lite Joey 1x CC BY-NC-SA 4.0 ESRGAN (lite) JPEG compression 2020-10-10 1x_DIV2K-Lite_80k.pth
NMKD Jaywreck3 & Jaywreck3-Soft (Lite) Nmkd 1x WTFPL ESRGAN Lite [nf=32 nb=12] JPEG compression 2020-10-13 1xESRGAN 1xESRGAN
MangaJPEG Bunzero++ 1x CC BY-NC-SA 4.0 ESRGAN JPEG 2021-06-17 Remove JPEG artifacts from manga without destroying screentone and other details none https://cdn.discordapp.com/attachments/579685650824036387/855269193762865152/Example_LQ_Input-comparison.png
Aliasing
Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
SS Anti Alias 9x BlueAmulet 1x CC BY-NC 4.0 ESRGAN Anti aliasing Dataset: Custom (9x Supersampling AA) none (no interpolation)
Anti Aliasing Twittman 1x CC BY-NC 4.0 ESRGAN Anti aliasing / Images with pixelated edges Dataset: Custom (?) none (no interpolation)
1x_GainRESV3 (Aggro,Natural,Passive) CF2lter 1x WTFPL ESRGAN Anti aliasing / Deblur 2022-02-26 To eliminate aliasing and general artifacts caused by not enough resolution while bringing out details Im stopping its training here because it's getting worse, i think of some aligment issues by game's rendering pipeline + downscaling... Dataset: 5K resolution shots from paladins rendered by 200% for hr and 37.5%(1080p) for lr then downscaled to 1080 BCGONE_DetailedV2 https://cdn.discordapp.com/attachments/579685650824036387/947132730313949184/cmp2.png https://cdn.discordapp.com/attachments/579685650824036387/947132731056324668/cmp.png
GIF
Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
DeGif nmkd 2x WTFPL ESRGAN GIF Restoration GIF Restoration 2xESRGAN
DDS (BC1/DXT1, BC3/DXT5 Compression)
Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_BC1-smooth2.pth BlueAmulet 1x CC BY-NC 4.0 ESRGAN BC1 Compression 2019-09-05 A model to help remove compression artifacts in BC1-BC3/DXT1-DXT5 compressed images (these all have color encoded the same way) none
x1_ITF_SkinDiffDDS_v1 intheflesh#3116 1x CC BY-NC 4.0 ESRGAN BC1 Compression 2022-08-24 Removes banding, blocking, dithering, aliasing, noise and color tint on DDS Compressed Skin Diffuse Textures.

This should work extremely well on most modern DDS compression types. The training set was compressed with BC3/DXT5, BC3/DXT5 Fast, BC2/DXT3, BC2/DXT3 Fast, and a small number of JPEG compressed images to cover outliers.

This model is trained to remove the slight green color tint that DDS compression tends to add to skin textures, so the model output will not match the original color tone of the input image. This is the desired result though, as DDS compression shifts the colors to a sickly green tint and this model corrects that to more natural color tones.

The training set included faces, body parts, eyes, mouths and hair in a variety of skin types and tones so it should work well on most related diffuse textures.

However it's not just limited to skin, many other images and textures can be cleaned with this model. Designed to be used as a first step cleaning pass before applying additional models after. Check out the other ITF Models.

1x_BC1-smooth2.pth https://imgsli.com/MTIyMzE5 https://imgsli.com/MTIyMzIw https://imgsli.com/MTIyMzIx
1x_DXTDecompressor_Source_V3-300000_G JosephtheKP 1x WTFPL ESRGAN DXT1 Compression 2021-12-21 Removing compression artifacts from DXT1 compressed textures. This model was created to remove DXT1 compression artifacts from textures imported into the Source Engine. Compressed textures in the engine sometimes have a green-tint which this model also corrects. The data for this model contained a good mix of diffuse textures and normal maps which means this model is pretty good at removing compression from normals as well. Creating this model was a real learning experience for me and I hope someone finds a good use for it. https://imgsli.com/ODcwNDg https://imgsli.com/ODcwNDk https://imgsli.com/ODcwNTA https://imgsli.com/ODcwNTE
1x_DEDXT CF2lter 1x WTFPL ESRGAN DXT compression 2022-02-26 To retain details while removing artifacts caused by dxt compression on textures https://cdn.discordapp.com/attachments/579685650824036387/947116367205793882/cmp.png
1x_artifacts_bc1_free_alsa.pth Alsa 1x CC BY-NC-SA 4.0 ESRGAN BC1 Compression BC1 take 2
1x_BCGone_Smooth_110000_G Mutin_Choler 1x WTFPL ESRGAN BC1 Compression 2020-08-02 Attempts to remove the damages done by BC1 compression. 1xESRGAN https://imgsli.com/MjU5MzQ/4/3
1x_BCGone-DetailedV2_40-60_115000_G Mutin_Choler 1x WTFPL ESRGAN BC1 Compression 2021-03-18 1xESRGAN
Dithering
Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


1x_artifacts_dithering_alsa.pth Alsa 1x CC BY-NC-SA 4.0 ESRGAN Dithering Dithered Images JPG (0-20%)
1x_DEDITHER_32_512_126900_G .pth Alsa 1x CC BY-NC-SA 4.0 ESRGAN Dithering JPG (0-20%)
4xFSDedither Jacob 4x GNU GPL3 ESRGAN Dithering 2020-01-29 For photos/realistic images, but worth trying on other images that have reduced colors and dithering along with fine details. Trained using the ESRGAN-FS code (https://github.com/ManuelFritsche/real-world-sr/tree/master/esrgan-fs/codes) for better details compared to plain ESRGAN. RRDB_ESRGAN_x4 https://imgsli.com/MTI1NTY/
4xFSDedither_Manga Jacob 4x GNU GPLv3 ESRGAN-FS Dithering 2020-04-21 Cartoons/pixel art/other non-realistic stuff with dithering RRDB_ESRGAN_x4.pth https://imgsli.com/MTQ3Nzc
4xFSDedither_Riven Jacob 4x GNU GPLv3 ESRGAN-FS Dithering 2020-07-11 Fine-tuned 4xFSDedither to upscale images from the game Riven, but should be better in general, particularly on ordered dithering. I adjusted the dataset to have a better variety of dithering parameters, and turned up the HFEN and pixel loss to get better details and color restoration with less noise. 4xESRGAN https://imgsli.com/MTg5NTM https://imgsli.com/MTg5NTQ https://imgsli.com/MTg5NTU https://i.imgur.com/j7Wtn0G.png
1x_DitherDeleter-Smooth_104000_G Mutin_Choler 1x WTFPL ESRGAN Dithering 2021-01-23 1xESRGAN https://imgsli.com/Mzg0MTk/1/2
1x_DitherDeleterV3-Smooth Mutin_Choler 1x WTFPL ESRGAN Dithering 2021-04-18 Attempts to remove the damages done by dithering. For this model, I downscaled all of the HR images to make their pixel size closer to 1000x1000 using the Box filter and then downscaled them again by 50% using the Point filter. Afterwards, I applied 32-bit Riemersma to every image in the dataset. 1xESRGAN https://imgsli.com/NTA5MjU/1/0

Blurring

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_ThePi7on-Solidd_Deborutify_UltraLite_260k_G ThePi7on 1x CC BY-NC-SA 4.0 ESRGAN Deblur 2021-05-09 Sharpening, line darkening and slight line thinning, specifically made for the Boruto anime.
1x_DeBLR BlackScout 1x Unknown or None ESRGAN Deblur 2020-02-19 General Deblurring
mdeblur LyonHrt 1x UNKNOWN ESRGAN Blurring Strong deblurring model
1x_PixelSharpen_100000 DinJerr 1x CC BY-NC-SA 4.0 ESRGAN Pixel Art 2021-03-19 Restores blurry/upscaled pixel art. 1xESRGAN https://imgsli.com/NDUxNDQ/5/4
1x_ArtClarity DinnJerr 1x WTFPL ESRGAN Pixel Art 2021-08-05 Texture retaining denoiser and sharpener for digital artwork 4xPSNR
1x_Fatality_DeBlur Twittman 1x CC BY-NC-SA 4.0 ESRGAN Deblur 2019-10-16 Dataset: Mix of anime, manga, and photos 1x_DeJpeg_Fatality_01_175000_G.pth
UnResize V3 (1x_UnResize_V3_200000_G.pth) Twittman 1x CC BY-NC-SA 4.0 (?) ESRGAN Deblur / Unresize Dataset: Anime, manga, and photos -

Purpose: Fix images that have been arbitrarily / poorly resized, such as non-integer nearest-neighbor upscaling/downscaling - Also acts as an image sharpener/deblur when used on slightly soft inputs

1x_UnResize_MKII_030000_G.pth
1x-Focus and 1x-Focus_Moderate Kim2091 1x CC BY-NC-SA 4.0 ESRGAN Deblur These models deblur most images. It was trained mostly on aniso2 and iso blurring (BSRGAN augmentation) with some gaussian mixed in. It performs well on most blurry images, but I'd recommend using something like Fatality_Deblur for very strong gaussian blur. Dataset: The UniScale Dataset + My Fabric dataset 1xESRGAN Sample 1 Sample 2 Sample 3 Sample 4
ReFocus V3 (1x_ReFocus_V3_140000_G.pth) Twittman 1x CC BY-NC-SA 4.0 ESRGAN Deblur / ReFocus 2022-03-12 Dataset: Photos, Anime, and manga -

Purpose: DeBlur, ReFocus, Sharpen Real life style images, but will work on Anime images too.

1x_Saiyajin_DeJPEG_300000_G.pth
ReFocus Cleanly (1x_ReFocus_Cleanly_100000_G.pth) Twittman 1x CC BY-NC-SA 4.0 ESRGAN Deblur / ReFocus 2022-01-21 Dataset: Anime, manga, and photos -

Purpose: DeBlur, ReFocus, Sharpen Manga, Anime and cartoon style images, but will work on real life images too.

1x_ReFocus_V3_110000_G.pth

Banding

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_N64clean BlueAmulet 1x CC BY-NC 4.0 ESRGAN Banding 2020-07-16 N64 textures use a color depth of 5-bits per channel, this model attempts to clean them, restoring smooth gradients in textures 1x_BC1-smooth2.pth
1x_Bandage-Smooth Mutin_Choler 1x WTFPL ESRGAN Banding 2021-04-16 Attempts to remove the damages done by color banding. For this model, I downscaled all of the HR images to make their pixel size closer to 1000x1000 using the Box filter and then downscaled them again by 50% using the Point filter. Afterwards, I applied 64-bit color banding to every image in the dataset. 1xESRGAN https://imgsli.com/NTA1NTk/1/0
Debandurh_FS Ultra-lite (1x_Debandurh_FS_lite_140000_G.pth) Twittman 1x CC BY-NC-SA 4.0 (?) ESRGAN Debanding images Dataset: Anime, real life, manga None

Halo Removal

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_DeEdge_105000_G Mutin_Choler 1x WTFPL ESRGAN Halo remover 2020-11-03 1xESRGAN https://imgsli.com/MjgxMDU/3/0

Noise

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
LADDIER1 Alexander Syring 4x Unknown ESRGAN Denoise 2019-11-13 Remove noise, grain, box blur, lens blur and gaussian blur and increase overall image quality. RRDB_ESRGAN_x4.pth
1x_NoiseToner-Poisson-Detailed_108000_G Mutin_Choler 1x WTFPL ESRGAN Noise remover 2021-10-08 Attempts to remove the damages done from noise. Successor of sorts to Noisetoner_Poisson_150000_G 1xESRGAN https://imgsli.com/NzQ5MzM/2/0
1x_NoiseToner_Poisson_150000_G Mutin_Choler 1x WTFPL ESRGAN Noise remover 2020-10-14 1xESRGAN https://imgsli.com/MjU1ODU/3/1
1x_NoiseToner_Uniform_100000_G Mutin_Choler 1x WTFPL ESRGAN Noise remover 2020-10-14 1xESRGAN https://imgsli.com/MjU1Nzg/2/1
1x_ISO_denoise_v1 Alpha 1x WTFPL ESRGAN denoise not on discord Remove high ISO noise 1xESRGAN
1x_ISO_denoise_v2 Alpha 1x WTFPL ESRGAN denoise not on discord Remove high ISO noise ISO denoise v1
Film-Degrainer_1-000 Tika 1x CC0 ESRGAN denoise not on discord Remove film grain/noise none


Oversharpening

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_DeSharpen Loinne 1x CC BY-NC-SA 4.0 ESRGAN Denoise 2019-06-03 Made for rare particular cases when the image was destroyed by applying noise, i.e. game textures or any badly exported photos. If your image does not have any oversharpening, it won't hurt them, leaving as is. In theory, this model knows when to activate and when to skip, also can successfully remove artifacts if only some parts of the image are oversharpened, for example in image consisting of several combined images, 1 of them with sharpen noise. 1st attempt on random sharpening with the same dataset at 200000 iterations, which was trained on non-random desharp model, total ~600000 iterations on 3 models.

DeToon

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


detoon LyonHrt ESRGAN Detooning 2019-06-24 A toon to realistic shading style model to wiki under drawings
1xDoubleDetoon Joey 1x CC BY-NC-SA ESRGAN Detooning 2020-07-31 An attempt to detoon images/drawings of people 1xESRGAN



Image De/Colorization

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
BS_Colorizer/Vapourizer BlackScout 1x GNU GPLv3 ESRGAN Colorization 2020-01-30 B/W | 100% Desaturated images. It mostly results in Blue and Yellow images with slight hints of Green, Orange and Magenta. You are free to use this as a pretrain to achieve better results. 1xESRGAN
1x_SpongeColor-Lite Joey 1x CC BY-NC-SA 4.0 ESRGAN (lite) [nf=32 nb=12] Colorization 2020-10-15 The first attempt at ESRGAN colorization that produces more than 2 colors. Doesn't work that great but it was a neat experiment. none
Deoldify Rastrum 4x CC0 ESRGAN Photos 2019-08-29 Old black and white photo restoration. Falcon Fanart

Stylization

Images

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
Rek's Effeks Photoanime v2 Rek 4x GNU GPL3 ESRGAN Stylization 2020-10-17 Photo stylization from JPEGs. Trained on images upscaled by ISO Denoise v2 -> DeJPEG Fatality PlusULTRA -> NMKD Yan2. Essentially a combination of that model chain into one. Use if you're looking for a stylized output, not photo quality. NMKD Yandere2
Ghibli_Grain.pth nonogamester#3975 1x CC BY-NC-SA 4.0 ESRGAN Realistic Ghibli Grain 2022-11-16 Text or Anime Attempt to get the nostalgic grain feel of classically animated Ghibli movies. Got the idea from researching the making of Ronin's :PKT_blue: 1x_AnimeFilmGrain28k. This was made running Nate video denoising with the preset "more denoising" on kiki's delivery service then overlaying it 66.65 percent over the original with DaVinci. On digital drawn anime it gives a slightly [10%] more organic/sharp feel to black lines. Matches well with content that already has a light digital grain. Run Twice for heavy grain. Eternal Thanks to all that indulge all my :psyduck: Wonderings :element_fire_neon: https://slow.pics/c/M6dNTu6G https://imgsli.com/MTM0Nzcw/9/8

Specialized Models

Text

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
2x_BSTexty BlackScout 2x GNU GPLv3 ESRGAN Text 2020-03-04 As the name might suggest, this model aims to upscale text with less distortion that other models. It seems to do a good job generally, but don't expect it to be a state of the art model that can upscale magazines and stuff. It makes things more readable but since it was train on B/W pictures it desaturates them. 2x_ESRGAN
4x-TextSharpV1 Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Text 2021-12-28 Text or Anime Interpolation between 4x-UltraSharp and 4x-TextSharp-v0.5. This is a duplicate listing of 4x-AnimeSharp https://media.discordapp.net/attachments/556604553424928768/925526934555881532/unknown.png https://cdn.discordapp.com/attachments/903415274521374750/925533775616696340/unknown.png


NMKD Typescale Nmkd 8x WTFPL ESRGAN Text 2020-11-04 Low-resolution text/typography and symbols 8xESRGAN

Inpainting

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_sudo_inpaint_PartialConv2D_424000_G.pth sudo rm -rf / --no-preserve-root#8353 1x CC BY-NC-SA 4.0 ESRGAN Inpainting 2020-12-05 Experimental PartialConv2D attempt to paint with ESRGAN. Took ~10.4 days of training on a P100 and around 1.5 months in total due to Colab limits. Not sure if I will continue training it since training is very slow, but may get better.. Warning: Result can vary with different tilesizes. Try not to tile your data. None Samples
1x_NMKD-YandereInpaint_375000_G.pth Nmkd 1x Unknown ESRGAN Inpainting Unknown None Samples

Fabric/Cloth

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
4x-Fabric and Fabric-Alt Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Fabric 2021-09-15 This model set upscales fabric or cloth textures (works on cats too!). The Alt model is just an earlier iteration version. It may work better on some images.The images need to be minimally compressed or passed through a decompression model first. It works with DDS compression though. 4xESRGAN Sample 1 Sample 2 Sample 3 Sample 4


Alphas

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
FireAlpha BlueAmulet 4x CC BY-NC 4.0 ESRGAN Alpha (4 channel) 2019-11-12 none
4x_1ch-Alpha-Lite Joey 4x CC BY-NC-SA 4.0 ESRGAN Alpha 2020-12-15 Obsoleted by Joey's Fork - Alpha channels of PNGs none
2x_Gen5-Alpha Joey 2x CC BY-NC-SA 4.0 ESRGAN Pixel Art with Tranparency / Alpha Channel 2021-02-03 4xFireAlpha
8x_Sphax-Alpha-NN Joey 8x CC BY-NC-SA 4.0 ESRGAN Pixel Art with Tranparency / Alpha Channel 2021-02-04 Replica of sphax with transparency 4xFireAlpha
4x_PocketMonsters-Alpha Joey 4x CC BY-NC-SA 4.0 ESRGAN Pixel Art with Tranparency / Alpha Channel 2021-02-04 upscaling pixel art with alpha channels that should perform better than any other. It should work well on both cartoon and 3D styled content. 4xFireAlpha
Skyrim Alpha Deorder 4x CC0 ESRGAN Pixel Art with Tranparency / Alpha Channel Dataset: Alpha Channels from Skyrim Unknown

CGI

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1xRedImage10000.pth 3majsie1995 1x CC BY-NC 4.0 ESRGAN Correct old color photos that are tinted red 2022-08-04 Correct old color photos that are tinted red. Model only tested in nature. 1xESRGAN https://i.imgur.com/5O74rRL.png
WaifuGAN_v3_30000 DinnJerr 4x CC BY-NC 4.0 ESRGAN CGI 2019-07-12 Upscaling CG-painted anime with variable outlines. Manga109v2.pth
2xBS_Wolly BlackScout 2x GNU GPLv3 ESRGAN CGI Animation 2020-02-07 Pixar Movies or Wall-E pictures/frames 2xESRGAN
4xSGI ChrisNonyminus 4x GNU GPL3 ESRGAN CGI 2020-12-17 Upscaling and dedithering pre-rendered sprites, images and textures made in the 90s. Basically vintage CGI. 4xESRGAN https://imgur.com/a/iwLTjeB
CGIMaster_v1 Madiator 2x GNU GPL3 ESRGAN CGI Animation 2021-01-14 mixed 3D/2D CGI animations on some 2D animations it might sharpen the edges. General usage is to upscale CGI animations compressed by YouTube. Sharp_Anime_v2 https://imgsli.com/Mzc1NTM

Luminance/Chroma

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


1x_BSLuma BlackScout 1x GNU GPLv3 ESRGAN Luma 2020-02-28 Fix Luminance issues?? LumaSharpen ESRGAN Edition? - This model mostly does what "Lumasharpen" algorithms do. It may help fixing images with Luminance images issues? Like old DVD rips? I am not sure, didn't test. none
1x_BSChroma BlackScout 1x GNU GPLv3 ESRGAN Sharpen 2020-02-29 ChromaSharpen - makes the colors slightly more vibrant with a sideeffect of possibly adding Chromatic Abberation to the image. I am not sure about the usage of this model on real case scenarios but anything blurry or with fuzzy colors could work none

Cats

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
Cat_Patch Twittman 4x CC BY-NC-SA 4.0 (?) ESRGAN Cats not on discord previous attempt

Coins

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
4x_Nickelfront BlackScout 4x GNU GPLv3 ESRGAN Coins 2020-02-13 Upscale coins. That's it. If you were mad at me because Nickelback doesn't make any sense. Now you have the perfect solution to your problems. If you want to upscale nickels or anything with similar texture made out of metal, now you can. It works pretty well for a joke. 4xESRGAN

Faces

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
4x_Faces_N_250000.pth Twittman 4x ESRGAN Faces 2019-07-18 Upscales images of faces of different scales, sometimes resulting in monsters. 4xESRGAN
4x_Faces_04_N_180000_G Twittman 4x ESRGAN Faces 2019-07-23 Upscale faces both pixelized and real 4x_Faces_N_250000.pth
Face-Ality V1 (4x_Fatality_Faces_310000_G.pth) Twittman 4x CC BY-NC-SA 4.0 (?) ESRGAN Faces Dataset: Custom (Faces) 4x_Faces_04_N_180000_G.pth
4x_SmolFace_200k DinJerr 4x CC BY-NC-SA 4.0 ESRGAN Art/People 2022-08-19 A sharp upscaler trained specifically for small sprite faces. Does not blend, so avoid using on painted/photo portraits unless you were trying to retain more of the outlines somehow. 4x_NMKD-UltraYandere_300k https://imgsli.com/MTIxNjU1/8/9
BigFace_v3, BigFace_V3_Blend, BigFace_V3_Clear DinJerr 4x CC BY-NC 4.0 ESRGAN Art/People 2020-07-22 4x_BigFace
TGHQFace8x tg 8x GNU GPL3 ESRGAN Faces 2019-11-27 Upscales blurry 128px faces, usefull for enhancing that someone in a small picture. 8xESRGAN https://imgsli.com/OTM1NQ
FArtFace DinJerr 4x CC BY-NC 4.0 ESRGAN Faces 2019-11-28
4x_BS_SbeveHarvey BlackScout 4x GNU GPLv3 ESRGAN Faces 2020-07-24 Upscale Steve Harvey, but maybe other things, somehow?? 4xESRGAN
Face Focus LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Faces not on discord Face De-blur - slightly out of focus / blurred images of faces. It is aimed at faces / hair 4xPSNR
TGHQFace8x Torrentguy 8x [https://www.gnu.org/licenses/gpl-3.0.en.html GNU GPLv3] ESRGAN Face Upscaling 8xESRGAN


Skin

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


x1_ITF_SkinDiffDetail_Lite_v1 intheflesh#3116 1x CC BY-NC-SA 4.0 ESRGAN Skin Upscaling Adds plausible high frequency detail and removes subtle blur. This is an early unfinished attempt at a x1 Lite model designed specifically for enhancing detail on skin diffuse textures of 3d characters. Even in its current state it works quite well.

Best suited for uncompressed or cleaned textures - otherwise it may just enhance any existing compression artefacts too. The training set included faces, body parts, eyes and hair in a variety of skin types and tones so it should work well on most related diffuse textures.

However it's not just limited to skin, many other images and textures can be enhanced with this model. The results are subtle, so run multiple times if desired.

50/50 Interpolation of DIV2K-Lite and SpongeBC1-Lite


Foliage/Ground

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_Plants_400000_G.pth Muf 1x Public Domain ESRGAN bad upscale 2021-06-23 Images of plants, trees or other foliage upscaled with Photoshop Preserve Details 2.0. Sharpens and "subdivides" details and noise so it doesn't look upscaled. 1x_NMKD-h264Texturize_500k.pth https://imgsli.com/NTg2MDk https://imgsli.com/NTg2MDg https://imgsli.com/NTg2MTA
Ground tldr_coder 4x WTFPL ESRGAN Upscales ground textures 2021-06-23 Dataset: Custom (Ground textures Google)

Video Games

Game Screenshots

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
Minepack BlackScout 4x CC BY-NC-SA ESRGAN Upscaling pack.png 2020-01-26 Upscales Minecraft screenshots by 4x. May suffer from haloing, weird patterns on blocks and JPEG-like artifacts. 4xESRGAN.pth
4x_Link Fielran#1024 4x GNU GPL3 ESRGAN Chainmail game textures. Alternatively, it can be used to turn plain images into chainmail. 2022-10-06 A highly generative model for chainmail game textures. Works fairly well on items without visible chainmail texture (ie just a grey area) or items with clear chainmail texture, less well on items with present but poorly defined chainmail texture. I'm not 100% happy with it but it is still an improvement over existing models in enough situations to be worth releasing. 4x_Nickelfront_14000G.pth https://cdn.discordapp.com/attachments/579685650824036387/1027685415756509265/4x_Link_500000_G-TestImage-0x9F6EFFE0.png https://cdn.discordapp.com/attachments/579685650824036387/1027685416180121722/4x_Link_500000_G-TestImage-CMMM_Knight_Templar_ID2_DiffuseMap.png https://cdn.discordapp.com/attachments/579685650824036387/1027685416528261203/4x_Link_500000_G-TestImage-Crest_Venetian_DiffuseMap.png
4x-FatePlus-lite Kim2091 4x CC BY-NC-SA 4.0 ESRGAN-lite Anime PSP games, Fate Extra 2022-07-03 This model was trained as a favor to Demon and the Fate Extra community. It leaves a nice grain on the images and upscales lines and details accurately without looking odd. This model works on most anime-style PSP games. Enjoy! It works best on content with dithering and quantization. NOTICE: I have included both NCNN and ONNX models to make upscaling easier if you rely on either of these. For NCNN, there's two versions. One is FP16 and the other is FP32. FP16 works best on RTX GPUs. Choose FP32 if in doubt about compatibility, or if FP16 doesn't work for you. To use ONNX, download chaiNNer and upscale through there with the ONNX nodes. 4x-AnimeSharp-lite
4x_GameAI_2.0 4x_GameAI_1.0 Tal 4x WTFPL ESRGAN PS2 textures, cartoonish and realistic game textures. 2022-02-27 This model is intended to mainly handle PS2 compression and a mixture of Realistic and cartoonish textures, it's not meant to be used for very low resolution textures such as item icons. Dataset: I used textures from, "(A Hat in Time, Kingdom Hearts 3 and World of Final Fantasy) for GameAI_2.0" "(Skyrim, The Witcher 3, Resident Evil 4, Final Fantasy XV and Tales of Vesperia ) for GameAI_1.0" Dataset_size: 21.9k to 70k textures , up to 1024x1024 per texture 4x_GameAI_1.0 https://cdn.discordapp.com/attachments/579685650824036387/947610558981632040/2ea384c8dc05593e-55207ec648ad52a4-00006213_2-comparison.png https://cdn.discordapp.com/attachments/579685650824036387/947610559497510922/65aac3d41ce0297c-5ff21b7095173a35-00006213_2-comparison.png https://cdn.discordapp.com/attachments/579685650824036387/947610560005033984/a6518186b4e88a86-ddbb8bd9666cd975-00006213_2-comparison.png https://cdn.discordapp.com/attachments/579685650824036387/947610560369950740/ae6ae340a472b13d-2ade3b7d0cf4a09f-00005dd3_2-comparison.png
4xPackCraft_v4 Joey 4x CC BY-NC-SA 4.0 ESRGAN Upscaling pack.png 2020-09-05 Designed to upscale one specific minecraft screenshot. Results on dissimilar screenshots may be poor. 4xESRGAN/previous versions https://cdn.discordapp.com/attachments/436411385673547787/750481717307113543/pack.png
MinecraftAlphaUpscaler with Good data Washed Up 4x GNU GPLv3 ESRGAN Upscaling pack.png 2020-01-29 Designed to upscale one specific minecraft screenshot. Results on dissimilar screenshots may be poor. RRDB_PNSR_x4.pth
ScreenBooster V2 BlackScout 4x CC BY-NC-SA ESRGAN CGI 2020-01-29 Game Screenshots none
BS_ScreenBooster_SPSR BlackScout 4x CC BY-NC-SA SPSR CGI 2020-07-18 This model is designed to upscale game screenshots (3D Games) by 4 times. The SPSR version is an improvement over the ESRGAN based V2. Screenbooster V2

Normal Map/Bump Map Generation

You may want to use this instead: https://github.com/JoeyBallentine/Material-Map-Generator

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_NormalMapGenerator-CX-Lite Joey 1x CC BY-NC-SA 4.0 ESRGAN Lite [nf=32 nb=12] Map Generation - normal maps 2020-11-04 Generating normal maps from textures 1x_DIV2K-Lite_450k.pth
1x_FrankenMapGenerator-CX-Lite Joey 1x CC BY-NC-SA 4.0 ESRGAN Lite [nf=32 nb=12] Map Generation - roughness and displacement maps 2020-11-07 This model generates "Franken Maps" (named after Frankenstein), which is a custom material map combination I made. Basically, the Red channel of RGB is just the texture converted to grayscale, the Green channel is the roughness map, and the Blue channel is the displacement map. I had to do this to get around the current limitation of CX loss where it requires a 3 channel output (otherwise I would have just made a 2 channel model, or separate single channel models). As of right now the channels need to be manually split from each other but I will be making a tool for doing this automatically in the coming days. 1x_DIV2K-Lite_450k.pth
1x_normals_generator_general_215k LyonHrt 1x CC BY-NC-SA 4.0 ESRGAN Map Generation - Normal Maps This model generates "Franken Maps" (named after Frankenstein), which is a custom material map combination I made. Basically, the Red channel of RGB is just the texture converted to grayscale, the Green channel is the roughness map, and the Blue channel is the displacement map. I had to do this to get around the current limitation of CX loss where it requires a 3 channel output (otherwise I would have just made a 2 channel model, or separate single channel models). As of right now the channels need to be manually split from each other but I will be making a tool for doing this automatically in the coming days. none (no interpolation)

Video Game Textures

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample


Fallout Weapons (Fallout 4 Weapons?) Bob 4x CC BY-NC-SA 4.0 ESRGAN Game textures 2019-08-05 Video game textures, mostly metal rusty, clean or painted Manga109Attempt
Fallout Weapons V2 Bob 4x CC BY-NC-SA 4.0 ESRGAN Game textures 2019-08-23 Video game textures, mostly metal rusty, clean or painted Fallout 4 Weapons
4x_ThiefGoldMod_100000 Akven 4x no idea ESRGAN Game textures 2020-05-15 Version of the previous model but based on Manga109 pretrained model and with slightly different dataset. Sometimes gives better results especially for wood and metal, sometimes worse. Sometime generates the same dotted artifacts on very bright/white images. 4x_Manga109Attempt https://imgsli.com/MTYzNjI
4x_ThiefGold_110000 Akven 4x no idea ESRGAN Game textures 2020-05-15 Various game textures. Primary wood, metal, stone RRDB_ESRGAN_x4 https://imgsli.com/MTYzNjI
4x-VolArt Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Game textures/Art 2021-09-18 This model upscales artwork for the game Volfoss (2001). The model peaked in quality very quickly. The NR model removes most noise, but has the downside of removing transparent portions. Use the main model in most cases. 4xESRGAN Sample 1 Sample 2
2xFaithfulSPSR Joey 2x CC BY-NC-SA SPSR Pixel art 2020-07-19 Mainly just a test for SPSR. Seems to work better than the original 2xFaithful32_1316 that I used as a pretrained, even though it uses the same dataset. 2xFaithful32_1316
2x_FakeFaith-Lite Joey 2x CC BY-NC-SA 4.0 ESRGAN (lite) Pixel Art 2020-10-12 An attempt at recreating the "faithful" style without using the faithful dataset -- aka keeping the "pixel art" style of pixel art. 2x_Faithful-Lite
2x_Faithful-Lite Joey 2x CC BY-NC-SA 4.0 ESRGAN (lite) Pixel Art 2020-10-12 A "lite" model version of my Faithful model none
4x_HDCube Venomalia 4x CC0-1.0 ESRGAN Gamecube and Wii textures. 2022-05-10 Gamecube and Wii textures (mainly DXT and 8bit color compression). Is good for preserving fine details without affecting the original style too much, it is not suitable for pixel art, small icons and text under 16 pixel. 4x_NMKD Siax https://imgsli.com/MTA3NDQ3 https://slow.pics/c/rmQsQ6yb
1x_DXTless_SourceEngine_170000_G Xeller 1x WTFPL ESRGAN DTX5 Compression 2021-04-10 This model is made for Source Engine textures. It tries to remove compression artifacts such as blockyness, discoloration, green tint. It does pretty well on a lot of things, realistic stuff as well, but it was mostly made to work on TF2 textures. It's made to keep as much detail as possible, without any unnecessary denoising/sharpening. Huge thanks to Twittman for assisting me along this journey. 1x_Saiyajin_DeJpeg_300000_G https://imgsli.com/NDk1NTk https://imgsli.com/NDk1NjA https://imgsli.com/NDk1NjE https://imgsli.com/NDk1NjI
Trixie LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Faces / Game Textures not on discord character textures for star wars games, including the heroes, rebels, sith and imperial. Plus a few main aliens…Why called trixie? Because jar jars big adventure would be too long of a name… This also provides good upscale for face textures for general purpose as well as basic star wars none
Skyrim Armory Alsa 4x CC BY-NC-SA 4.0 ESRGAN Game textures/equipment not on discord Manga109Attempt
Skyrim Misc Deorder 4x CC0 ESRGAN Game textures not on discord Skyrim Diffuse Textures ?
Skyrim Wood Laeris 4x ESRGAN Game textures not on discord Wood ?
4x-SkyrimTexV2.1 and V2_Fabric Kim2091 4x CC BY-NC-SA 4.0 ESRGAN Game textures 2021-09-25 This model set upscales most Skyrim textures. I hope it helps 🙂 The base model (2.1) works well on stone, wood, metals, and most other textures. The Fabric model is a 50/50 interpolation with a stronger iteration of this model and my Fabric model. As you can tell by the name, it's intended for Fabric textures. If you want to upscale Alpha textures, use a model dedicated to it. There's an INFO file in the folder, it just explains the extra models 4xESRGAN Sample 1 Sample 2 Sample 3 Sample 4
Forest LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Game textures not on discord Wood / Leaves none
Morrowind Mixed mkultra 4x CC BY-NC-SA 4.0 ESRGAN Game textures not on discord Morrowind Mod Textures 4xESRGAN
Morrowind 2.0 mkultra 4x CC BY-NC-SA 4.0 ESRGAN Game textures not on discord Morrowind Mod Textures 4xESRGAN
Map LyonHrt 4x CC BY-NC-SA 4.0 ESRGAN Game textures not on discord Map / Old Paper with text none

Video Restoration

Video Compression

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x_cinepak_200000 Twittman 1x CC BY-NC-SA 4.0 (?) ESRGAN Compression 2019-07-03 Removal of Compression such as Cinepak, msvideo1 and Roq none
1x_cinepak_alt.pth buildist 1x GNU GPLv3 ESRGAN Compression Cinepak (only) none
DeBink v4/v5/v6 Kim2091 1x CC BY-NC-SA 4.0 ESRGAN Compression 8.6.21 This model removes early 2000s Bink and other compression artifacts. Works well on almost any image or video compression type. none
DeBink_Lite Kim2091 1x CC BY-NC-SA 4.0 ESRGAN-lite Compression 2021-10-30 This model was trained on lossless video frames of Metal Arms: Glitch in the System, compressed with Bink V1 for the LR frames. It's a lot more efficient than my first DeBink model, and also has less artifacts. It's not quite as robust, but the compression is barely noticeable in videos after processing with this. none Video Sample
1x-RoQ_nRoll Kim2091 1x CC BY-NC-SA 4.0 ESRGAN Compression 2021-11-21 This model decompresses images and video compressed using RoQ. Config and presets will be added when Mega decides to let me use their site! none https://imgsli.com/ODI5Njg

https://cdn.discordapp.com/attachments/911838608879677440/912086092755374200/unknown.png https://cdn.discordapp.com/attachments/911838608879677440/912084886792314880/unknown.png https://cdn.discordapp.com/attachments/911838608879677440/912085213859954788/unknown.png https://cdn.discordapp.com/attachments/911838608879677440/912087766366564372/unknown.png

1x_DeRoqBeta-lite Kim2091 1x CC BY-NC-SA 4.0 ESRGAN Compression 2022-07-26 Incomplete lite model to remove ROQ compression https://cdn.discordapp.com/attachments/579685650824036387/1001671437293207672/1658887461.0522466.png
NMKD h264Texturize Nmkd 1x WTFPL ESRGAN Texturizing 2020-10-20 Tries to reverse heavy h264 compression. Fails. Can be used to texturize images though. 4x ESRGAN
1x_Filmify4K_v2_325000_G.pth Muf 1x Public Domain ESRGAN artifact 2021-07-19 This model attempts to make films upscaled to 4K with Topaz Gaia-HQ look more natural and filmic. It sharpens, adds film grain, and smooths out small artefacts from the upscaling process. I recommend adding a tiny amount of grain to the input to seed the model (you can do this in VEAI), otherwise the film grain will remain static across frames that don't move much. Pretrain model used with permission to relicense from Twittman. 1x_UnResize_V3_110000_G.pth https://imgsli.com/NjE5MTE https://imgsli.com/NjE5MTI https://imgsli.com/NjE5MTM https://imgsli.com/NjE5MTQ
DeIndeo (mirror) DeIndeo Wild West Quest#7975 4x WTFPL ESRGAN Indeo Compression Artifacts Dataset: Custom 4xESRGAN

VHS Tapes

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1xBaldrickVHSFix_180000_G_V0.2 NimRodZorg 1x CC BY-NC-SA 4.0 ESRGAN VHS 2021-03-03 Fixing minor VHS Chroma and Pattern Noise - NOTE: only works on deinterlaced sources 1xESRGAN
ToonVHS Redslam 1x CC BY-NC-SA 4.0 ESRGAN VHS 2022-02-06 Best when used on cartoons, it can work on anime. Due to the dataset it does struggle a bit with orange colors and grainy dark spots. This model is meant to be used to clean up the image before using it on a 2x or 4x model. 1xESRGAN Examples of ToonVHS + other models Examples of ToonVHS by itself
2x_VHS-upscale-and-denoise_Film_477000_G Itewreed 2x CC BY-NC-SA 4.0 ESRGAN VHS 2021-03-28 VHS captures of Film material, but may work on VHS recorded native SD-TV material as well. Also useable for cleaned up source material none
VHS-Sharpen-1x RTX 2080 ti hoarder 1x CC BY-NC-SA 4.0 ESRGAN VHS 2021-03-19 Make old VHS footage crispy. This model will not work on video and images with noticeable JPEG/Video compression artifacts, noticeable interlacing or haloing, heavy tape distortion/artifacts and scenes with tons of detail. For best results, use a downscaled HD capture of the VHS tape you intend to use it on. 1xESRGAN https://imgsli.com/NDUxNDY/2/3

Model Collections

Model Name Author Scale License Architecture Purpose Date Posted Description sample
OptimusPrimal's Collection OptimusPrimal Various WTFPL ESRGAN Collection 2022-02-28 I'm dumping all my models here. (I moved my favorites to a favorite folder). Most of these models work well on DVD resolution animation sources. These are all ESRGAN models. Everything is with the WTFPL license. So do whatever you want with them. Credit is not needed, but would be nice if you use them to make something. There's a few that are interpolations with other user's publicly available models, or interpolations of other user's publicly available models (so technically, these are not mine), so credit goes out to them, the only ones I can remember off the top of my head is @twittman with the Fatality model series, which are great, and @cd VSGAN && poetry install with the Sol Levante/American Dad2 models, and if you do 50/50 with those 2 models, you get a really good model that I called Soladad.
Kim's Collection Kim2091 Various CC BY-NC-SA 4.0 Various (ESRGAN and ESRGAN-lite) Collection 2022-02-28 My (Kim's) main folder for hosting models. Every model of mine on this database links to subsections of this folder. If a link is broken or you just want to see nearly all of the models I've trained, look here.
UltraMix Collection Kim2091 Various CC BY-NC-SA 4.0 Various (ESRGAN and ESRGAN-lite) Collection 2022-06-28 This is a mixture of models based around UltraSharp and my other available models. These are usually interpolations that have separate but very helpful uses. As an example, UltraMix_Restore is a combination of UltraSharp and UniScale_Restore, and is great for video game textures.
NMKD's Collection NMKD Various Various Various (ESRGAN and ESRGAN-lite?) Collection 2022-02-28 NMKD's main folder for hosting models. Nearly every one of NMKD's models posted on this wiki link to it.
Joey's ESRGAN Bot Collection Various Various Various Various Collection 2022-08-04 Here is a dump of all the ESRGAN-Bot models. As it was a un-curated collection, many of the models here were never officially released, but I do not know exactly which ones. If one of your models is included here and you would like me to remove it, please let me know.


Pretrained Models

Looking for official models? Look here: https://upscale.wiki/wiki/Official_Research_Models

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1xESRGAN victorca25 1x Apache License 2.0 ESRGAN Pretrained 2019-07-06 RRDB_ESRGAN_x4.pth
2xESRGAN victorca25 2x Apache License 2.0 ESRGAN Pretrained 2019-07-06 RRDB_ESRGAN_x4.pth
4xESRGAN xinntao 4x Apache License 2.0 ESRGAN Pretrained 2019-07-06 RRDB_ESRGAN_x4.pth
8xESRGAN victorca25 8x Apache License 2.0 ESRGAN Pretrained 2019-07-06 RRDB_ESRGAN_x4.pth
16xESRGAN victorca25 16x Apache License 2.0 ESRGAN Pretrained 2019-07-06 RRDB_ESRGAN_x4.pth
1x PSNR Pretrained Model BlueAmulet 1x Apache License 2.0 (ESRGAN's license) ESRGAN Pretrained 2020-04-20 The original RRDB_PSNR_x4.pth model converted to 1x, 2x, 8x and 16x scales, intended to be used as pretrained models for new models at those scales. These are compatible with victor's 4xESRGAN.pth conversions RRDB_PSNR_x4.pth
2x PSNR Pretrained Model BlueAmulet 2x Apache License 2.0 (ESRGAN's license) ESRGAN Pretrained 2020-04-20 The original RRDB_PSNR_x4.pth model converted to 1x, 2x, 8x and 16x scales, intended to be used as pretrained models for new models at those scales. These are compatible with victor's 4xESRGAN.pth conversions RRDB_PSNR_x4.pth
4x PSNR Pretrained Model BlueAmulet 4x Apache License 2.0 (ESRGAN's license) ESRGAN Pretrained 2020-04-20 The original RRDB_PSNR_x4.pth model converted to 1x, 2x, 8x and 16x scales, intended to be used as pretrained models for new models at those scales. These are compatible with victor's 4xESRGAN.pth conversions RRDB_PSNR_x4.pth
8x PSNR Pretrained Model BlueAmulet 8x Apache License 2.0 (ESRGAN's license) ESRGAN Pretrained 2020-04-20 The original RRDB_PSNR_x4.pth model converted to 1x, 2x, 8x and 16x scales, intended to be used as pretrained models for new models at those scales. These are compatible with victor's 4xESRGAN.pth conversions RRDB_PSNR_x4.pth
16x PSNR Pretrained Model BlueAmulet 16x Apache License 2.0 (ESRGAN's license) ESRGAN Pretrained 2020-04-20 The original RRDB_PSNR_x4.pth model converted to 1x, 2x, 8x and 16x scales, intended to be used as pretrained models for new models at those scales. These are compatible with victor's 4xESRGAN.pth conversions RRDB_PSNR_x4.pth
Compact Pretrained Models Zarxrax 1x-4x WTFPL Real-ESRGAN "compact" Pretrained 2022-07-31 This is a collection of pretrained models for Real-ESRGAN's Compact architecture. There are 1x, 2x, and 4x models, as well as 1x and 2x "UltraCompact" and "SubCompact" models (think of these as the equivalent to ESRGAN "lite" models). By using these are pretrains for your models, you can ensure that your models are able to be interpolated with other Compact models that were trained from these. These pretrains are compatible with most existing compact models.

Pretrained Discriminators

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
1x Pretrained Discriminator Pack Joey 1x CC BY-NC-SA 4.0 VGG Pretrained discriminators 2020-10-26 Most of these are my spongebob dataset so they will be more useful for cartoons but I did include my original faithful model discriminator. none
2x Pretrained Discriminator Pack Joey 2x CC BY-NC-SA 4.0 VGG Pretrained discriminators 2020-10-26 Most of these are my spongebob dataset so they will be more useful for cartoons but I did include my original faithful model discriminator. none
4x Pretrained Discriminator Pack Joey 4x CC BY-NC-SA 4.0 VGG Pretrained discriminators 2020-10-26 Most of these are my spongebob dataset so they will be more useful for cartoons but I did include my original faithful model discriminator. none
8x Pretrained Discriminator Pack Joey 8x CC BY-NC-SA 4.0 VGG Pretrained discriminators 2020-10-26 Most of these are my spongebob dataset so they will be more useful for cartoons but I did include my original faithful model discriminator. none
4x_RRDB-G_ResNet-D (Both G and D) Joey 4x CC BY-NC-SA 4.0 ESRGAN (G) / ResNet (D) Pretrained 2021-01-17 Clean bicubic downscales / Pretrained models (G/D) RRDB_ESRGAN_x4_old_arch.pth

Real-ESRGAN Compact Custom Models

These are the new, small and fast Real-ESRGAN "compact" models. These are different from normal ESRGAN or Real-ESRGAN models. They can be used with chaiNNer.

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
Compact Pretrained Models Zarxrax 1x-4x WTFPL Real-ESRGAN "compact" Pretrained 2022-07-31 This is a collection of pretrained models for Real-ESRGAN's Compact architecture. There are 1x, 2x, and 4x models, as well as 1x and 2x "UltraCompact" and "SubCompact" models (think of these as the equivalent to ESRGAN "lite" models). By using these are pretrains for your models, you can ensure that your models are able to be interpolated with other Compact models that were trained from these. These pretrains are compatible with most existing compact models.
sudo_UltraCompact_2x_1.121.175_G.pth (pcloud) / sudo_UltraCompact_2x_1.121.175_G.pth (mediafire) sudo 2x CC BY-NC-SA 4.0 Compact Realtime animation restauration and doing stuff like deblur and compression artefact removal 2022-05-29 My first attempt to make a REALTIME 2x upscaling model while also applying teacher student learning. It beats Anime4k in every way. These benchmarks use a 3060ti and it shows that everything better than a 3060ti should be able to handle 1080p input if you create engine files and use my TensorRT code. You can see in the readme how to convert onnx files into engines. The 2 right bars compare normal Compact2 and Ultracompact in speed, the 2 on the left showcase older apis I used which isn't too important for this showcase. To use this, you need to use my code which is https://github.com/styler00dollar/VSGAN-tensorrt-docker. If you use Manjaro, it is also possible to pipe the data stream directly into mpv, so you can watch it in a video player without rendering a video. Yeah the model does seem a little noisy if you zoom in a lot, but don't forget that the model itself is only 1.2mb. I think it does quite well. I still try to improve on fast models, but this is good enough to share as a first model. Plz don't steal without credits, k thx. RealESRGANv2-animevideo-xsx2.pth (Teacher: RealESRGANv2-animevideo-xsx2.pth) https://cdn.discordapp.com/attachments/579685650824036387/980550564725260358/compare.png https://cdn.discordapp.com/attachments/579685650824036387/980550565354410066/unknown.png https://cdn.discordapp.com/attachments/579685650824036387/980550565723529226/compare2.webp
2x_DigitalFlim_SubCompact Zarxrax 2x WTFPL Compact

[nf:24 nc:8]

Animation 2022-05-14 This was trained on the dataset that OptimusPrimal used for his DigitalFilm models. This model cleans up the image removing some noise while upscaling. This model is very fast, running around 200x the speed of a standard ESRGAN model on my system. This is primarily a proof of concept for how fast Real-ESRGAN models can get while still producing nice results. https://imgsli.com/MTA3ODYx

https://imgsli.com/MTA3ODY0

https://imgsli.com/MTA3ODY2

https://imgsli.com/MTA3ODY3

1x_HurrDeblur_SubCompact Zarxrax 1x WTFPL Compact

[nf:24 nc:8]

Animation 2022-06-04 This is a sharpening/deblurring model for anime video. It was created with three goals in mind:
  • Be blazing fast
  • Try to avoid enhancing noise and compression artifacts
  • Try to avoid over-enhancing intentionally blurred parts of the image

Despite that last point, this is not intended to be used on modern anime which makes heavy of use depth-of-field effects.

https://imgsli.com/MTEwOTg4

https://imgsli.com/MTEwOTY2

https://imgsli.com/MTEwOTY5

https://imgsli.com/MTEwOTcx

1x_AnimeUndeint_Compact Zarxrax 1x WTFPL Compact Animation 2022-06-05 This model corrects jagged lines on animation that has been deinterlaced. It handles simple line doubling, line interpolation, and even Yadif-style artifacts. It can also handle sources that were resized after deinterlacing, for example resizing from ntsc to pal resolutions. If a source has been upscaled after deinterlacing, it will need to be downsized before applying this model. https://imgsli.com/MTExMTE0

https://imgsli.com/MTExMTE1

https://imgsli.com/MTExMTE2

https://imgsli.com/MTExMTE3

1x_BleedOut_Compact Zarxrax 1x WTFPL Compact Animation 2022-07-25 This model helps repair color bleed and heavy chroma noise that may be present on some older footage, particularly that which was recorded on VHS. It also cleans up rainbows if they are present. https://imgsli.com/MTE4MjEz

https://imgsli.com/MTE4MjE4

https://imgsli.com/MTE4MjE5

2x / 4x-anifilm_compact Kim2091 2x and 4x CC BY-NC-SA 4.0 Compact Animation 2022-08-02 This model is based on a private model by @eula 5600x 3070 named 4x_eula_anifilm_v1_225k. He sent me a copy of the model, and I decided to train a compact model based on it with his permission. This model seems to fix the majority of the issues the original model had while being far faster, it's just a tiny bit softer in some images.

The dataset consists of Dragon Ball movies converted to YUV24 with @sgdisk --zap-all /dev/sda's help to reduce artifacts, then upscaled with ArtClarity and eula_anifilm. LRs are the original frames right from DVD. As a result, this model corrects some color space issues. The 2x model's HRs were downscaled by 50% with Lanczos.

The 2x and 4x models are pretty close in output despite being trained separately. The 2x model is a bit softer overall.

The models in the Real-ESRGAN Compatible folder are the original output from Real-ESRGANs training code for compatibility reasons.

4x_Compact_Pretrain.pth 2x Comparison:

Sample 1 Slider Comparison 1

4x Comparison: Sample 2 Slider Comparison 2

Official Models

To cut down on the length of this page and reduce confusion, these models have been moved here: https://upscale.wiki/wiki/Official_Research_Models This page contains models such as: ESRGAN, BSRGAN, Real-ESRGAN, and more.

SOFVSR Models

These models are for SOFVSR, they will not work in Cupscale or the main ESRGAN forks. You need this to use it: https://github.com/JoeyBallentine/Video-Inference

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
3x_Video_TSSM Joey 3x CC BY-NC-SA 4.0 SOFVSR Animation 2020-11-17 none
3x_Video_TSSM_RRDB Joey 3x CC BY-NC-SA 4.0 SOFVSR RRDB / VESRGAN Animation 2020-11-25 none
2x_SBS11-RRDB Joey 2x CC BY-NC-SA 4.0 SOFVSR RRDB / SOFVESRGAN Animation 2020-12-03 none
4x_REDSVAL-7f-RRDB-Lite (both G and D) Joey 4x CC BY-NC-SA 4.0 SOFVSR RRDB / SOFVESRGAN Lite [nf=32, nb=12] IRL videos 2020-12-03 none
SOFVSR_REDS_F3_V1 Sunseille 4x WTFPL RRDB IRL videos 2021-02-26 upscale IRL videos. Use not recommended by creator none https://cdn.discordapp.com/attachments/547949806761410560/813134983236419664/100k_iter.mp4
2x_VimeoScale_Unet Sazoji 2x CC BY-SA 4.0 SOFVSR-RRDB (light arch, 3 frames, nf=32, nb=12) Upscaling video content 2021-09-09 This model is meant to surpass VEAI 2x while being efficient to run quickly with fp16. The real-esrgan/BSRGAN augmentation and Unet should help with videos where the resolution is not ideal and can reconstruct details without effecting blurs in most cases. This model SHOULD run faster than real-esrgan while matching the resolving and enabling some multiframe feature extraction. No major denoising/compression/blurring effects (or artifacts) should be found. Self-Trained Base for the Unet finalisation with vgg_fea discrim Sample 1 Sample 2 Sample 3
4x_VimeoScale Sazoji 4x CC BY-SA 4.0 SOFVSR-RRDB (5 frames) Upscaling IRL video content 2021-10-08 This model is one of the longest I have trained and tuned, includes some noise training but this is not intended for deblocking/decompression. Combined with https://github.com/JoeyBallentine/Video-Inference 's fp16 mode, this should handle most SD and 720p content at or faster than ESRGAN, with a significant bump in quality. Self-Trained Base for the Unet finalisation with vgg_fea discrim Sample 1 Sample 2 Sample 3 Sample 4

Pretrained SOFVSR Models

Model Name Author Scale License Architecture Purpose (short) Date Posted Purpose (Full) Pretrained_Model_G sample
video_G victorca25 4x Unknown SOFVSR net Trained on REDS' size dataset none
VESRGAN_G victorca25 4x Unknown RRDB Trained on REDS' size dataset none

RIFE Models

Model Name Author Scale License Architecture Purpose Date Posted Description Pretrained_Model_G sample
sudo_rife4_269.662_testV1_scale1.pth sudo 2x CC BY-NC-SA 4.0 RIFE Animation interpolation 2022-06-25 I never really mentioned it in model-releases prior since I think not too many care about interpolation here, but I trained a rife4 model for animation a some months ago, which is better than rife4 and rife4.2 imo. Thought I should also mention it here as well. I also converted it into ncnn. (Nihuis rife ncnn models are only exported with the fastest mode and not the best quality. I exported ncnn models for the most important quality settings. Due to different export/quality settings, there are multiple models. For that reason alone, my ncnn models are much better too, since nihui only exported the fastest one.) My https://github.com/styler00dollar/VSGAN-tensorrt-docker also has the rife ncnn extention, which can use VMAF, dedup, scene detection and so on, which I would recommend. My models are in that extention as well, just select model 10, 11 or 12 and use the dev docker. That test video is done with 2x framerate, enbemble True and FastMode False, combined with scene detection and dedup stuff, tta False. Towards the best quality rife can do. Plz don't steal without credits, k thx. https://cdn.discordapp.com/attachments/579685650824036387/990345004260151296/ngnl_sudorife.mp4

CAIN Models

CAIN is a video frame rate interpolation AI. These models are intended for use with CAIN only. These CANNOT be used with Cupscale or ESRGAN, you need this: https://gitlab.com/hubert.sontowski2007/cainapp
Model Name Author Scale License Architecture Purpose Date Posted Description Pretrained_Model_G sample
explodV1 hubert 2x BSD-3-Clause CAIN YUV Animethemes 2022-07-29 I think one of sharpest models for anime

Plz don't steal without credits, k thx.

https://cdn.discordapp.com/attachments/579685650824036387/1002663424695750686/output.mp4
190562_vimeo_enchanced.pth hubert 2x MIT CAIN RL videos - interpolation 2021-07-13
cainREALTIME hubert 2x MIT CAIN (CAIN 1 Group) RL videos - interpolation 2021-07-13 Trained mostly on abba music videos
cainliteanime.pth hubert 2x MIT CAIN Anime - interpolation 2021-08-06 Sample
rvpV1 (pcloud) rvpV1 (mediafire) sudo rm -rf / --no-preserve-root#8353 2x CC BY-NC-SA 4.0 CAIN Video Frame Interpolation for anime openings 2021-09-14 My ~~first~~ (ok technically second) attempt to create a video frame interpolation model and I like how it turned out. To use it, you can either use https://gitlab.com/hubert.sontowski2007/cainapp, https://github.com/styler00dollar/Colab-CAIN or the bot in [Game Upscale] (model called rvpv1 there, just use --model rvpv1). Some demo videos in pcloud, but you need to download them. The web player seems to playback in low fps. The architecture is mainly the same to the original CAIN, but i modified the padding to be zero padding instead. ``.pt`` means JIT model, `.pth`, means normal pytorch model. Architecture file is in pcloud as well. And no, no cupscale or flowframes. Dataset: Modified Animeinterp dataset https://files.catbox.moe/xiq9vi.mp4
cvpv6.pth hubert 2x MIT CAIN Anime - interpolation 2021-09-21

Waifu2x Models

Some helpful links for Waifu2x, rewritten in PyTorch. The repos contain the pretrained models themselves as well.