search

volotat / SD-CN-Animation

This script allows to automate video stylization task using StableDiffusion and ControlNet.
MIT License
808 stars 63 forks source link

Txt2video seems to get stuck after random amount of frames generated #55

Open sgsavu opened 1 year ago

sgsavu commented 1 year ago

It ocurred to me in 3 instances where I was trying to generate max length videos. Stopping randonmly without error message in the console at 104, 123, 88. All with different settings (max fps, length, etc)

I'll try to have a look through the code soon if we can't seem to diagnose this.

sgsavu commented 1 year ago

To bump to this I think the issue might be caused by cuda out of memory or something related. I have an RTX 2060 SUPER with 8GB. And I can see that while rendering frames it gets close to max but doesn't hit it. Might that the process crashes internally but no error is displayed ?

memewizards commented 1 year ago

My render was getting stuck at frames prepared: 11 / 88; Frames processed: 5 / 88.

I'm running 2 x RTX 4090.

I don't see any errors in the console.

I think this was caused by interrupting the output while it was processing.

Twice I deleted the outputs and restarted the runpod pod. Got back into the jupyter notebook and rebooted stable diffusion webui.

On the second time, using a new mp4 video input I got a fully processed MP4 output. I'm doing a second render now without a restart and it appears to be working correctly.