I’m using the WanGP11.31 version of LTX2 on NVIDIA 3070 GPUs with 8 GB VRAM. On a single machine, generating a 10-second video takes about 48 minutes. I have 10 such machines, each with 8 GB VRAM. My assumption is that, if I could split the frame generation across all 10 machines, with each machine generating part of the 10 -second video, the total time could be reduced to around 4.8 minutes. Is there a recommended way to distribute frame generation across multiple machines to achieve this kind of speed-up?
I’m using the WanGP11.31 version of LTX2 on NVIDIA 3070 GPUs with 8 GB VRAM. On a single machine, generating a 10-second video takes about 48 minutes. I have 10 such machines, each with 8 GB VRAM. My assumption is that, if I could split the frame generation across all 10 machines, with each machine generating part of the 10 -second video, the total time could be reduced to around 4.8 minutes. Is there a recommended way to distribute frame generation across multiple machines to achieve this kind of speed-up?