search

thygate / stable-diffusion-webui-depthmap-script

High Resolution Depth Maps for Stable Diffusion WebUI
MIT License
1.73k stars 159 forks source link

[FEATURE REQUEST] Batch processing for videos? #420

Open eyeEmotion opened 8 months ago

eyeEmotion commented 8 months ago

Hi,

is there a reason why you can't batch process video's? I have to cut a 2 hour movie into 3 minute-parts, even 2 minute-parts when using Depth Anything (which I finally got working and the results are amazing for what I need it for). And it takes around 30 to 40 minutes for generating, computing and saving a Depthmap video file. And then I have to manually add the next video. So I have to constantly be around to start the next clip and I have around 60 of them. Because it also consumes a lot of my RAM, there is not much else I can do on my computer while it's processing. So for me it would be ideal to line up a few videos so it can do its thing during the night, when I'm sleeping.

As the elements are already there, wouldn't this be straightforward to implement? And ofcourse make sure the generated depthmap videos don't get overwritten by the next in line. I think that can be addressed by saving the file for example as "dpt_" + filename of the source video.

Would help me a lot, as I already have a lot of manual cutting and merging to do because of the limits with RAM-Memory.

Cheers.

petermg commented 8 months ago

This sounds like an amazing idea!

semjon00 commented 5 months ago

The proper fix would be to make video generation streamable - it converts some frames (maybe, like, 100), then saves some frames, appending them to some file.