quick question before i download
do i need to adjust the gguf node by updating any pr or outside updates to get the nodes to register the format? or is it the same functionality as wan in that itll translate the same architecture to the nodes?
Not really, you can experiment, but in the repository there are some examples that currently already work well
on rtx 3060 getting OOM on q4km gguf model...
on rtx 3060 getting OOM on q4km gguf model...
lower res
on rtx 3060 getting OOM on q4km gguf model...
lower res
384x384. Is that low enough?
on rtx 3060 getting OOM on q4km gguf model...
lower res
384x384. Is that low enough?
well that and youd probably want to chunk the segments of seconds down to maybe 3 or 4 seconds maximum and it should help. maybe stick to like 61 seconds a clip and just use a video slice node to chunk specific sections of the reference video until you have your full output then edit them together in capcut or something?
are you doing work or just testing it?
to test if it works at all for you jsut run 21 frames at 240 by 240 to test. if that works slowly boost everything up 1 part at a time. increase frame count before anything in testing as thatll be the hardest thing to overcome. res can be buffed with upscalers post render. its time consuming but itll help you find your limits.