Topaz Models are generally better, but updates to the models take longer. Allowing users to use open source or third party models will expand the use for the apps.
Especially in relation to animated videos. Gaia CG does an ok job, but a standalone model would do better. Allowing outside models specially trained for certain reasons would open Topaz up to focusing on improvements to the app itself.
Is there any kind of standardized format for video enhancement models?
Most of the open models i’ve seen are based off PyTorch. Topaz claims to have a proprietary system for their models or training. But my guess it is also based on PyTorch with some modifications.
If it was based on PyTorch, I think the number of complaints about TVAI not using all of the available hardware would be nonexistent.
Just a hunch.
TVAI uses Openvino.
