vadim epstein

thanks for your appreciation! on my experience, `--notext` worked somehow [sometimes] with short prompts, but didn't bring any noticeable improvements for longer text strings, which is way more common use...

text prompts parsing is finally in place, both for aphantasia [still] and illustrip [motion], local and colabs. `--notext` option therefore is eliminated everywhere, as it hardly compatible with multiple weighted...

thank you, glad to hear you liked it. alas, i've got zero experience with deepspeed yet, so can't help with it. i'll leave this issue open, in the hope it...

i do not think i'd invest more time in VQGAN methods: there are tons of various flavours for it on the internet. i've put my own VQGAN implementation here just...

@pidginred fixed sigma worked on my side for any resolution, including 1024x1024. it's not the reason of your problems.

@pidginred same as yours, 0.0625. but i also resize driving_video, not only source_image (which i see you don't).

@pidginred i've used it for rather artistic purposes (applying to face-alike imagery), so cannot confirm 100%. it definitely behaved very similar with 1024 and 256 resolutions, though. speaking animation quality,...

@zpeiguo that project is unreleased yet, sorry. and this topic is about high res images. check other issues for 'normality' of movements.

i'm facing the same problem, anybody any idea/solution?

num_workers=0 fixed it (1 didn't help)