📺 AniSD Release

AniSD is a suite of 12 specialized SISR models trained to restore and upscale standard definition digital anime from ~2000s and onwards, including both both WEB and DVD releases. Faithfulness to the source and natural-looking output are the guiding principles behind the training of the AniSD models. This means avoiding oversharpened output (which can look especially absurd on standard definition sources), minimizing upscaling artifacts, retaining the natural detail of the source and of course, fixing the standard range of issues found in many DVD/WEB release (chroma issues, compression, haloing/ringing, blur, dotcrawl, banding etc.). Refer to the attached infographic for a quick breakdown of the available models, and refer to the wiki page for an in-depth description of the models available.

The AniSD models are intended to be used with TRT inference (except for the CRAFT models, which don't support TRT). The SwinIR and DAT2 models in particular see significant speed uplifts from using TRT inference. For further details on how to use the models with TRT, refer to this wiki page.

Quickstart: I recommend 2x_AniSD_G6i1_SPAN_215K and 2x_AniSD_AC_G6i2b_SPAN_190K as your general-purpose models to try first. If you need something, heavier, try 2x_AniSD_AC_G6i2b_SwinIR_117500.

• Download all models (recommended)
• Download only PTH models
• Download only ONNX models

Upscaling Models

• AniSD: These are the baseline AniSD models. They are intended for "newer" standard definition sources, including WEB releases. 📌 Sample Comparisons
• AniSD Augmented Clean (AniSD AC): As the name suggested, the AniSD AC models are trained to deal with stronger degradations common in older DVDs. 📌 Sample Comparisons
• AniSD Deep Clean (AniSD DC): The final stand, so to speak. If a source isn't properly cleaned up by AniSD or AniSD AC, that's when you use AniSD DC. 📌 Sample Comparisons

Utility Models

• AniSD PostScale (AniSD PS): "Neutral" scaling that doesn't modify the base image. Use after running an upscaling model to hit a target resolution. 📌 Sample Comparisons (AniSD DC -> AniSD PS)
• AniSD DeBleed (AniSD DB): This 1x model is trained to deal with chroma bleed, and will help clean up a source before you run one of the upscaling models. The model isn't as potent as Zarxrax's Bleedout model, but also won't cause the drastic color shifting of Bleedout. 📌 Sample Comparisons

Credits 💖

• pwnsweet and Kuronoe for their contributions to the dataset
• @pifroggi for his ONNX conversion script, which enables SwinIR conversion
• Zarxrax for his chroma degradation script used to train AniSD DB
• @Kim2091's dataset destroyer, used for AniSD DB and almost certainly in other models as well
• NovaZone for convincing me to give SPAN another try
• and the numerous other members of the community that contributed to this process!

4️⃣🇰 Ani4K v2 Release

[Gen 6, Compact & UltraCompact] Ani4K v2, as the successor to the original Ani4K, retains its predecessor's fantastic detail retention, depth of field preservation and faithfulness to the original source. As the name suggests, the model is targeted at modern anime, ranging from high-quality Bluray to crappy WEB releases, for upscaling either to 2K or 4K.

• Image Comparisons: General Comparison

FAQ

• How does v2 differ from v1? I'm so glad you asked! A shortcoming of the v1 models is that the model really struggled on sources which were poorly mastered. This is unfortunately still very common even with modern anime. v2 is far more capable of dealing with such sources. 
• How does Ani4K v2 differ from JaNai v3? Which one should I pick? JaNai v3 is a fantastic model, and shares many of the fundamental training objectives behind Ani4K (DOF preservation, faithfulness, etc.). I'd say that the primary difference is one of training philosophy-- JaNai seeks to render the source as if it was originally mastered in 4K, whereas Ani4K seeks to produce an upscale that is as close as possible to the source (while cleaning up any issues). Long story short, test both models and see which you prefer.
• What model versions are there? Ani4K comes in Compact and UltraCompact flavors. Compact is of course a standard option. UltraCompact provides a noticeable performance uplift, without too much impact on quality. I ultimately did not train a SuperUltraCompact variant as I felt the hit to the model quality was far too significant.

OpenProteus

HD/FHD Live Action TV and Cinema - Open Proteus [Compact]

[Compact] Topaz Video AI is generally considered amongst the industry leaders of in upscaling movies and TV, however it sits behind a significant paywall. Open Proteus is a free alternative to Topaz's Proteus AI, and intended to upscale high-quality and low-noise HD sources to 4K. While I won't claim this is superior to Topaz Proteus by most metrics (which would be a near-impossible feat on open-source architectures anyway), Open Proteus produces output which is highly faithful, without causing oversharpening or upscaling artifacts.

• Image Comparisons: General Comparison

🥈 AniScale 2 - Second Batch of Models!

AniScale 2 has now received ESRGAN, ESRGAN Lite, DITN and SwinIR versions! For a breakdown of the new model versions, see the wiki page. The TL;DR is that the OmniSR version is still king, but these new models should provide additional variety to play with.

All model downloads can be found here.

🌟 AniScale 2 & AniScale 2 Refiner Release

[Gen 6, OmniSR, Compact, etc.] The successor and replacement to the original AniScale, and a substantial upgrade in nearly every respect. AniScale 2 is a versatile and faithful anime model trained for use on a variety of post ~2000 sources. Superb blur and depth of field handling, thorough WEB and DVD compression repair, and pleasing line art refinement are the hallmarks of AniScale 2.

• While AniScale 2 is trained first and foremost as an OmniSR model, AniScale 2 is also intended to be a platform to explore multiple SISR archs. For starters, I've trained OmniSR and Compact versions, but more will come.

• AniScale 2 also comes with a "refiner" model, creatively named AniScale 2 Refiner (AS2R). AS2R is a 1x Compact model (for maximum speed) trained to supplement and increase the versatility of AniScale 2, without making the base model excessively aggressive. The Refiner is focused on providing light sharpening, fixing line art and line thinning, depending on whether you run the model before or after upscaling with AniScale 2.

• For more information, please visit the Github wiki page.

• Image Comparisons:

  • Blur Comparison
  • General Comparison

• Detail Retention: ⭐⭐⭐⭐

• Compression Cleanup: ⭐⭐⭐⭐⭐

🎉2x Ani4K Compact Release

[Gen 5, Compact] As the name suggests, Ani4K is trained to create natural-looking 2K/4K upscales from 720p/1080p sources. In particular, the model has superb detail retention.

• Image Comparisons:
  • Blur Comparison
  • General Comparison
• Detail Retention: ⭐⭐⭐⭐⭐
• Compression Cleanup: ⭐⭐(will deal with compression in HD/FHD BD and WEB sources without a hitch, but may be less suited for DVDs-- use Anima or AniDVD for DVD sources)