Verarbeitung #374: Replace pydub for preview generation - collecting_society - C3S Tracker

Verarbeitung #374

Replace pydub for preview generation

Added by Alexander Blum about 7 years ago. Updated over 4 years ago.

Status:

Neu

Start date:

Priority:

Normal

Due date:

Assignee:

Thomas Mielke

% Done:

Category:

Estimated time:

5.00 h

Target version:

Repertoire 3) Testing phase III

Description

As we will definitly have the usecase of audio files with more than 2 channels and as pydub is not capable of handling this due to constraints in the python wave library, we need to replace it.

Maybe a combination of direct calls to ffmpeg for compression/conversion and sox for the preview composition (fadein/crossfades/fadeout) could be reasonable.

History

#1 Updated by Thomas Mielke about 7 years ago

Assignee changed from Thomas Mielke to Alexander Blum

Another option would be to first create a temporary mono/16bit/11.025Hz wav file using ffmpeg or sox. Then we use this to

a) cut out the excerpt
b) create the preview and
c) create the fingerprint.

This should notably increase processing speed. I suggest we change the preview format from 8.000 to 11.025 Hz to avoid a resampling step.

#2 Updated by Alexander Blum about 7 years ago

Assignee changed from Alexander Blum to Thomas Mielke

Thomas Mielke wrote:

first create a temporary mono/16bit/11.025Hz wav

good idea.

using ffmpeg or sox

then ffmpeg. sox is currently not installed.

preview format from 8.000 to 11.025 Hz to avoid a resampling step.

right now it's 16000
if the quality is acceptable, then I'm fine, but it could happen, that we need a bit more to comfort the ears of the web users.
the estimate of the preview size should also be included in Estimates.