create file format tag --fileformat for converting audio downloads #7
Conversation
…ned format with ffmpeg
caitriggs
changed the title
There was a problem hiding this comment.
Great job!
I want to keep this package as lightweight as possible so depending on ffmpeg is a little scary but I also see the benefit of getting mp3:s, Ogg:s or whatever out instead of a mix of formats.
Could you add a brief test of the audio conversion module (e.g https://github.com/caitriggs/audio-scraper/blob/fileformat/tests/test_audioconvert.py) so we can make sure it works?
audioscrape/__main__.py
Outdated
| """Scrape various websites for audio.""" | ||
| youtube.scrape(query, include, exclude, quiet, overwrite) | ||
| soundcloud.scrape(query, include, exclude, quiet, overwrite) | ||
| # create subdirectory for converted audio files if --fileformat tag set |
There was a problem hiding this comment.
Minor nitpick but capitalize and have normal sentences with periods for code comments, please. 🐼
There was a problem hiding this comment.
This is your show, so no problem there! Will do 👍
audioscrape/audioconvert.py
Outdated
| '-i', "./{}".format(file), # input file to convert | ||
| '-f', '{}'.format(fileformat), # output fileformat type | ||
| '-ac', '1', # mono channel | ||
| '-ar', '16000', # sampling rate 16000Hz |
There was a problem hiding this comment.
Should 16 kHz mono be the default or is 44100 Hz / 16-bit stereo more common? What are the intended use cases of the audio conversion?
There was a problem hiding this comment.
44100 Hz stereo is better quality, but the specific application I'm using your scraper for is for ASR model training which are generally built for 16000 Hz mono channel wav files (e.g. Bai Du's DeepSpeech). I should probably re-tool it to allow user defined settings, but admit I just added a quick and dirty hard coded solution for my needs
audioscrape/soundcloud.py
Outdated
| Appears that soundcloud streams are mp3 only, so we'll | ||
| convert original mp3 to user defined file format with ffmpeg. | ||
| ''' | ||
| # Convert to fileformat using ffmpeg |
There was a problem hiding this comment.
Condense the two comments into one.
audioscrape/youtube.py
Outdated
| itself until Stream.download(), we must convert | ||
| the audio after download with ffmpeg. | ||
| ''' | ||
| # Convert to fileformat using ffmpeg |
There was a problem hiding this comment.
Condense the two comments into one.
Yea, I totally understand keeping the dependency footprint small, but for me it's really convenient to build file conversion into the scraping process. Many of us in the data science community scrape data for a specific purpose (e.g. all input files have to be 16kHz mono wave files). ffmpeg is a pretty well established standard for audio/video encoding/decoding so thought it was safe library choice. Especially since if you're working with audio files in the first place, you're already likely to have ffmpeg installed.
Yup, shouldn't be an issue 👍 |
|
Closing because stale. |
Addresses feature request #6