search

Aubrey-ao / HumanBehaviorAnimation

172 stars 14 forks source link

why the length of the input audio, the output motion, and the rendered result are not match #11

Open Droliven opened 11 months ago

Droliven commented 11 months ago

I input the audio: TrinitySpeech_Gesture_I_GENEA_Challenge_2020/Test_Data/Audio/TestSeq001.wav, which has 1min 58 seconds.

after run generate.py, get motion TestSeq001.bvh, which has 2350 Frames, Frame Time: 0.050000

then import the bvh motion into the blender, it shows a 10s video, (25fps, 250frames).

My qeestion here is why the length of the input audio, the output motion, and the rendered result are not match

Aubrey-ao commented 11 months ago

I check the number of sample proint of "TestSeq001.wav" is 5640000 (at 48kHz), so if resample it into 20fps. the number of frames is 2350 (= 5640000 / 48kHz * 20fps).

Therefore, the length of the input audio matches with the output motion length.

As for the Blender Importing, I think maybe u forget to choose the "Update Scene Duration" setting.

image