View Single Post
Old 12th March 2011, 10:46   #1  |  Link
ar
Registered User
 
Join Date: Feb 2009
Posts: 16
Resampling audio for PAL Telecine

Task.
I got some FILM material (23.976 fps): hd-video with audio. I want to make DVD from this. I decided to use PAL, so i need to make telecine conversation.

Details.
For video it is no big deal: just to assume new frame rate. But for audio this task should be a bit complicated, because i need to convert sample rate or use some TimeStretch-filter.
To keep it simple i decided to use sample rate conversion wich lead to little speed-up and corresponding pitch-effect. Input have 48kHz audio and 24000/1001 fps video, output audio should be 48kHz to, but output video should be 25 fps. So the sample rate conversation scheme should be:
  1. Assume new sr for input audio samples 48000/(24000/1001)*25=50050 (Hz);
  2. Convert sample rate back to 48kHz.

I've done this with commands:
Code:
ffmpeg -i ed-cm-st-16bit.wav\
       -acodec pcm_f32le -ac 2 -f f32le pipe:1 |\
ffmpeg -ar 50050 -ac 2 -acodec pcm_f32le -f f32le -i pipe:0\
       -acodec pcm_f32le -ac 2 -f wav f32le_st.wav &&\
fsrctool -q push -i f32le_st.wav -o outst.wav -r 48000
Question.
I've used FSRC-filter (libfsrc) due to it's outstanding quality. Is there someone who know wich SRC-filter ffmpeg utilises? The manual says, it's «high quality polyphase filter». Is this filter homegrown or it is from libsrc (secret-rabbit-code)? Also, originaly i used to use SSRC-filter from Naoki Shibata.

PS:
Comparison of SRC-filters at http://src.infinitewave.ca/
ar is offline   Reply With Quote