Open whmc76 opened 2 months ago
So far I couldn't find anyone that is willing to create the custom node for this project, which is very weird. I am using the command-line to get this project to work at the moment.
There is a 5 min loading duration and a 12 min inference duration for each video. Have to batch everything together. A custom node will save about 5 min each time we change the seed number as the models have been loaded. The inference duration could be reduced using a GPU with 16gb vram. Mine is only 8gb.
And obviously, having a custom node will save a lot of disk space, since you don't need to setup a specific conda environment for it. I found a tutorial on writing a custom node, so I will try to make one myself.
This is really strange, I think it's even necessary in all video flows, looking forward to your project, even the simplest implementation is fine
There is one right now by Smthemex. https://github.com/smthemex/ComfyUI_FoleyCrafter
Seems to be slower than the command-line.
So far I couldn't find anyone that is willing to create the custom node for this project, which is very weird. I am using the command-line to get this project to work at the moment.
There is a 5 min loading duration and a 12 min inference duration for each video. Have to batch everything together. A custom node will save about 5 min each time we change the seed number as the models have been loaded. The inference duration could be reduced using a GPU with 16gb vram. Mine is only 8gb.
And obviously, having a custom node will save a lot of disk space, since you don't need to setup a specific conda environment for it. I found a tutorial on writing a custom node, so I will try to make one myself.