No rendered video in inference_result folder (Linux)

[aroncserveni]

Every frame is synthesized, process is complete but there's no rendered video in the ./asserts/inference_result folder.

Do you have an idea why it doesn't produce the .mp4 result?

[Inferencer]

Possible issue with ffmpeg ensure its detected in your env
I hid the ffmpeg print results so that's why your not seeing an error
while the env is activated you can check if ffmpeg is correctly detected with the following command
ffmpeg -version

[aroncserveni]

Thank you, ffmpeg was indeed not correctly installed/detected. Now it works.

I have another question. The downloaded result from ./asserts/inference_result is of a much lower quality than the original video I uploaded. Is there a possibility to increase the output quality?

[Inferencer]

Yeah comes down to either how torch is installed or which version I think, I was messing around with it a couple of weeks ago and got the audio model to use GPU but it decreased the speed of the frame inference by a lot, resolution quality did return, I will do a pip list tomorrow when I'm on my PC and see what I did.

[aroncserveni]

Thank you, I appreciate it

[aroncserveni]

Yeah comes down to either how torch is installed or which version I think, I was messing around with it a couple of weeks ago and got the audio model to use GPU but it decreased the speed of the frame inference by a lot, resolution quality did return, I will do a pip list tomorrow when I'm on my PC and see what I did.

Have you had a chance to make a pip list?

[Inferencer]

appears to be torch 1.12
I must have changed my cuda version but wont be able to tell which version due to more changes since then, whatever works with torch 11.2 i suppose https://pytorch.org/get-started/previous-versions/

[aroncserveni]

appears to be torch 1.12 I must have changed my cuda version but wont be able to tell which version due to more changes since then, whatever works with torch 11.2 i suppose https://pytorch.org/get-started/previous-versions/

Thank you! The following seems to have fixed it.

CUDA 11.6
torch==1.12.1
torchvision==0.13.1
torchaudio==0.12.1

When could we expect V2 to be released?