Feature Request: Dedicated AI Model for Anime Upscaling and Enhancement

[sinan.saymaz]

Hey there, Topaz team! I’ve been meaning to reach out about this for a while now. Topaz has gained a lot of popularity for its outstanding enhancement capabilities, especially in the realm of anime editing. In fact, about 80% of anime editors are familiar with Topaz, often using the Proteus model to elevate their work to the highest quality possible.

Here’s the thing: Given Topaz’s strong presence in anime editing, many editors like myself are wondering why there isn’t a dedicated AI model specifically tailored for anime. Models such as Real ESRGAN, Real CUGAN, and Waifu2x have set a high standard in this niche, and having a similar dedicated model from Topaz would undoubtedly delight anime editors. It would not only encourage more usage of Topaz but also lead to even better results, further solidifying Topaz’s position as a leader in enhancing anime quality.

I believe developing such a model could open up new possibilities and attract more anime enthusiasts to explore Topaz’s capabilities further. Looking forward to hearing your thoughts on this!

[kyle.topazlabs]

The Gaia model currently available was trained on animated and computer-generated content and can be used for animated videos.

[sinan.saymaz]

The Gaia model doesn’t improve quality at all. In fact, it tends to make it worse. Here is the proof of it:

[sinan.saymaz]

Only on the upscaled version it makes it a little bit better

But this is a huge diffrent as if you would use Models such as Real ESRGAN, Real CUGAN, and Waifu2x

[william.doane]

That’s super interesting as recently I am working on making screen savers for a friend of mines coffee shop. I made him a bunch of custom wall papers that are like outdoors lofi stuff that’s not copy written. It works Shockingly well for anime. Here I am thinking im onto something new lolol. It’s definitely extremely helpful. I can see a lot of studios re-mastering the old stuff with this kind of tech