Support longer length videos
This workflow is amazing. Best results I have got from any locally runnable free system.
But, any ways to handle longer length videos? Using a longer video starts the workflow OK, but then eventually gives a OOM error.
You need to find out the limit of your system, then segment the video and do them in batches. You can join them after they are processed. For example, if your system can only handle 30 secs of video and you need to process a 2 mins video, then process the video 30 secs at a time in 4 batches. Join them up after it's done.
The length of the video you can process before you get OOM is also affected by the resolution.
If your workflow handled the split->process->join for any video it would be more than a very cool demo.
The workflow I made is a very basic framework to get started. ComfyUI users can modify it for their own purposes, segmenting and joining the videos can easily be done. I leave it up to the users how they want to implement it.
"easily be done"? Can you do it? Or point me to the relevant nodes or an example?
Another similar upscaling workflow is FlashVSR that would also be awesome to patch with a split/process/rejoin system.
I think you should look at the workflow for WAN SVI.
WAN can only do 5 secs video because that's how long it's coherence last. The SVI workflow segmented the video into batches of 5 secs then joined them together for the final result. Using the same looping system, you can implement it on my workflow and lora. It will be difficult if you are not familiar with ComfyUI, please get someone who knows how to use it.
OK, if you do find it easy and get some time free, please do.
Otherwise if anyone else sees this topic and can do it, please do and let me know.
I will leave it at that.