It would be great if we could train a face from a collection of images to keep the consistency of the face on lower quality videos. My main issue I’ve had is Iris sometimes changes the face completely between frames. If you could have a baseline to say “this face looks like this” then maybe it would keep things more consistent when upscaling.
Training is part of the model creation process. The app isn’t capable of doing any “training,” and if it was our computers probably wouldn’t be capable of running it.
Suggestion is not about training the model, but about using reference images to guide the enhancement process. I have a collection of old high-quality photos that could serve as a reference for what the faces should look like. This could potentially help maintain consistency in the appearance of faces when upscaling lower quality videos. Is there a way to incorporate this approach in Iris?
That’s what training is. You feed the ai data and it uses that data as examples of desired results. The app is not capable of such “teaching” from user entered data.
Many bad actors could train AI for deepfakes which Topaz definitely does not want to be associated with. Could be helpful for a few people but generally not worth the headache for Topaz
