How to reduce batch size when using video reference on minimax h3 ref2va?
retained text
When using a video reference it says “Not enough GPU memory. Try reducing image resolution or batch size and run again” I already reduce the mp to 0.4 for a 10s video. It can work if I reduce to 0.3mp but then video quality will be really bad. How can I fix this by having longer generation time but not losing too much quality? on 3090 Ti 24Gb vram/64gb ram
document fetched 2026-09-20 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1wko8nb/how_to_reduce_batch_size_when_using_video