171–180 of 320 posts
It’s astounding that 2 sentences generated this. (I used text-to-image and the prompt for a space marine in power armour produced something amazing with no extra tweaks required).
Earlier quoted context omitted.
All for suggestions! We've tried white noise as well, but it only works on plain talking samples (not music, for example). My guess is that the most robust solution will come from updating how it's trained.
What if you train it to hold the last frame on silence (or quiet noise)?
It completely falls apart on longer videos for me, unusable over 10 seconds.
One small issue I've encountered is that sometimes images remain completely static. Seems to happen when the audio is short - 3 to 5 seconds long.
Looks like too much Italian training data
[flagged]
[flagged]
We are big fans of Hedra. Do you know if they've publicly commented on their model architecture? As far as we know, our particular choice of an end-to-end diffusion + transformer is novel. We don't know what Hedra is doing. It could be the approach EMO has taken ( https://humanaigc.github.io/emote-portrait-alive/ ) or VASA ( https://www.microsoft.com/en-us/research/project/vasa-1/ ) or Loopy Avatar ( https://loopyava…