sstzal/DFRF

Result video shortage

Vadim2S opened this issue · 9 comments

Vadim2S commented

I am try few experiments with cnn.mp4 video from you examples. Original has 33 second and produce 840 frames for torso and head.
I am run rendering.sh script with different models and get output videos with 421 frames and 16 second only.

What is it?

s1752729916 commented

hi, i have the same problem and how you solved it?

flyingshan commented

I have the same problem, and I used ffmpeg to slow down the result video 2x, and it looked correct, but I dont know why...

  • 👍1
Teoge commented

I think the video is split into training and validation sets in half. And rendering.sh only renders the validation set.

AIMads commented

@Teoge I think so too, I have no clue how it mixes it with the audio.

hylq66 commented

@Teoge I think so too, I have no clue how it mixes it with the audio.

Do you know where the audio for the resulting video is stored, the second half of the original audio?

hylq66 commented

I think the video is split into training and validation sets in half. And rendering.sh only renders the validation set.

Do you know where the audio for the resulting video is stored, the second half of the original audio?

  • 👍1
cucdengjunli commented

same question

sstzal commented

I think the video is split into training and validation sets in half. And rendering.sh only renders the validation set.

It is right.

sstzal commented

I think the video is split into training and validation sets in half. And rendering.sh only renders the validation set.

Do you know where the audio for the resulting video is stored, the second half of the original audio?

Yes. You just need to combine the generated video with the driven audio you used.