AI Upscaling

By Zarxrax

Updated 6/27/2025

Overview

This guide provides an overview on how to use community-made AI upscaling models to improve the quality of anime videos. These models can generally provide the best overall quality when it comes to upscaling and cleaning up anime video.

A model can simply be thought of as a blob of computer code that has been trained to modify an image in a particular way. There are a lot of different community-made models to choose from. Some models don’t even do upscaling–they might do other tasks like remove certain types of artifacts from the image. Lots of different models can be found at OpenModelDB. I will recommend a few models to start with later in this guide.

There are a few important things to know about models. First, they have a fixed upscaling factor--typically 2x or 4x. So for example, if you have a video file that is 852x480, using a 2x model on it will result in a file that is 1704x960. In order to get it to a standard resolution like 1920x1080, you will need to follow up with a normal resize operation after upscaling.

Each model is trained upon one of many different architectures. When browsing models on OpenModelDB, it will tell you which architecture each model uses. All that you really need to know is that some architectures are faster, while others are slower but potentially higher quality. Some of the fastest architectures which work well for video are Compact and SPAN.  Slightly slower are RealPLKSR, ESRGAN, or anything with "lite" or "light" in the model name. Anything else might be quite slow, but you can try them yourself to see.

The other thing to know is that models can come in different file formats. The most common format is .pth, however .onnx is also common. Models can be easily converted between these two formats, so don't worry if a particular model is not in the format that you need.

A final thing to be aware of is that upscaling can potentially be quite slow, and will utilize your GPU. Nvidia GPUs perform best, but AMD Radeon GPUs are also alright. If your computer doesn't have a dedicated GPU, then upscaling will likely be too slow to be feasible.

When and Why to Upscale

Upscaling models are not magic. The age-old rule of "garbage in, garbage out" applies. It is ALWAYS better to start with the highest quality source footage possible, rather than using sub-par source video and hoping to improve it through upscaling. Proper filtering through AviSynth or VapourSynth before upscaling can work wonders.

You also don't want to go overboard with upscaling. A lot of proprietary software or services might make ridiculous claims about being able to upscale 16x or up to 8k! The fact is, upscaling a DVD source by 2x is typically all that needs to be done. If your source is already 1080p HD, then you probably don't need to upscale it any more at all. Think about this: when you upscale by 2x, you are generating 4 pixels for every one pixel that was in the original image. At 4x, you are generating 16 pixels for every one pixel from the original image! The more you upscale, all you really stand to gain is sharper lines. The model doesn't know what tiny details are supposed to look like, so the more you upscale, the more weird any tiny details will start to become. Videos that have been upscaled too much are usually pretty obvious--they have sharp lines but the details are all messed up, or many of the details have just been wiped out.

You might be surprised to learn that most anime is not even produced at full 1080p. It's true--most anime is actually drawn at a resolution somewhere between 720p and 1080p, then they just scale it up. You probably can't even notice that, can you? I just mention this to caution against trying to upscale to 4k or higher, because it simply isn't necessary in most cases.

The real use case for upscaling, in my opinion, is taking standard definition video and upscaling it to HD. This can often give a real visual improvement beyond what you could get from just AviSynth or VapourSynth.

One other thing I will mention, is that if you plan to zoom in on any clips while making your AMV, upscaling those clips can be a big help to prevent them from looking blurry! 

Upscaling Software

Here are two different applications that I recommend for upscaling video. You only need to choose one!

VideoJaNai - This is the recommended upscaling application for windows users. Its quite easy to use, and the process is pretty self-explanatory. It does require you to use models that are saved in the .onnx format.

Real Video Enhancer - This one is also quite good, and it works on multiple operating systems. It can also interpolate videos to increase the framerate, though you typically shouldn't do that! Downloaded models should be placed in the "custom_models" folder.

Power users may also be interested in vs-mlrt or avs-mlrt, which can let you upscale directly in VapourSynth or AviSynth, respectively.

Why not Topaz?

Topaz is a popular (and expensive) proprietary application for upscaling video. For live-action video, topaz is great! However, as a commercial product, they target their models towards mainstream uses, NOT for people trying to upscale anime videos. While many people still use topaz for anime, it's not optimal and you can get much better results with models that have actually been trained on anime!

Recommended Models

You can find upscaling models at OpenModelDB. Here are a few that I recommend.

(2x) AniSD Suite - This is a collection of models for dealing with DVD sources. Because there are so many models, choosing which one to use can be slightly confusing, so I highly recommend reading over this wiki page which goes into detail about the different models.

(2x) AnimeJaNai V3 - This model is for upscaling anime from 1080p to 4k. The effect is quite subtle, so it is really only effective on relatively high quality sources. It should be included with both VideoJaNai and Real Video Enhancer.

(2x) LD-Anime_Compact - Upscales while fixing numerous video problems, including: noise/grain, compression artifacts, rainbows, dot crawl, halos and color bleed. Can over-smooth some textures though. 

Of course, if none of these seem to work well for your source, feel free to try out any other models in the model database to see if there are any that work better for you.

Converting Models

If you have a model that is in .pth format and you need to convert it to .onnx, you can use this simple online converter: https://huggingface.co/spaces/Zarxrax/sisr2onnx