Comment by giancarlostoro

yt-dlp https://github.com/yt-dlp/yt-dlp lets you list and download different formats offered by YoTube for videos.

You could try downloading different audio formats and see if any of the available ones contain non-garbled audio.

For example:

  yt-dlp -F "https://youtube.com/watch?v=a28_aXgrgXE"

Output:

  [youtube] Extracting URL: https://youtube.com/watch?v=a28_aXgrgXE
  [youtube] a28_aXgrgXE: Downloading webpage
  [youtube] a28_aXgrgXE: Downloading ios player API JSON
  [youtube] a28_aXgrgXE: Downloading player 62ccfae7
  WARNING: [youtube] a28_aXgrgXE: nsig extraction failed: Some formats may be missing
           n = l9bNLKrDBBdCJtknGqU ; player = https://www.youtube.com/s/player/62ccfae7/player_ias.vflset/en_US/base.js
  WARNING: [youtube] a28_aXgrgXE: nsig extraction failed: Some formats may be missing
           n = Z-MlQqu4ClRjI62sqw_ ; player = https://www.youtube.com/s/player/62ccfae7/player_ias.vflset/en_US/base.js
  [youtube] a28_aXgrgXE: Downloading m3u8 information
  [youtube] a28_aXgrgXE: Downloading MPD manifest
  [info] Available formats for a28_aXgrgXE:
  ID       EXT   RESOLUTION FPS CH │   FILESIZE   TBR PROTO │ VCODEC        VBR ACODEC     ABR ASR MORE INFO
  ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
  sb2      mhtml 48x27        1    │                  mhtml │ images                               storyboard
  sb1      mhtml 33x45        1    │                  mhtml │ images                               storyboard
  sb0      mhtml 67x90        1    │                  mhtml │ images                               storyboard
  233      mp4   audio only        │                  m3u8  │ audio only        unknown            Default
  234      mp4   audio only        │                  m3u8  │ audio only        unknown            Default
  139-dash m4a   audio only        │  427.60KiB   48k dash  │ audio only        mp4a.40.5  48k 22k DASH audio, m4a_dash
  139      m4a   audio only      1 │  427.60KiB   48k https │ audio only        mp4a.40.5  48k 22k low, m4a_dash
  140-dash m4a   audio only        │  853.17KiB   96k dash  │ audio only        mp4a.40.2  96k 44k DASH audio, m4a_dash
  251      webm  audio only        │  851.41KiB   99k dash  │ audio only        opus       99k 48k DASH audio, webm_dash
  140      m4a   audio only      1 │  853.17KiB   95k https │ audio only        mp4a.40.2  95k 44k medium, m4a_dash
  160      mp4   108x144     30    │ ~962.40KiB  108k dash  │ avc1.4d400b  108k video only         DASH video, mp4_dash
  269      mp4   108x144     30    │ ~  1.43MiB  164k m3u8  │ avc1.4D400B  164k video only
  278      webm  108x144     30    │ ~846.56KiB   95k dash  │ vp9           95k video only         DASH video, webm_dash
  133      mp4   180x240     30    │ ~  2.11MiB  242k dash  │ avc1.4d400c  242k video only         DASH video, mp4_dash
  229      mp4   180x240     30    │ ~  2.65MiB  305k m3u8  │ avc1.4D400C  305k video only
  242      webm  180x240     30    │ ~  1.92MiB  220k dash  │ vp9          220k video only         DASH video, webm_dash
  134-dash mp4   270x360     30    │    3.32MiB  459k dash  │ avc1.4d4014  459k video only         DASH video, mp4_dash
  230      mp4   270x360     30    │ ~  5.07MiB  583k m3u8  │ avc1.4D4014  583k video only
  134      mp4   270x360     30    │    3.32MiB  380k https │ avc1.4D4014  380k video only         240p, mp4_dash
  243      webm  270x360     30    │ ~  3.53MiB  406k dash  │ vp9          406k video only         DASH video, webm_dash
  135      mp4   360x480     30    │ ~ 10.05MiB 1155k dash  │ avc1.4d400b 1155k video only         DASH video, mp4_dash
  231      mp4   360x480     30    │ ~ 11.43MiB 1313k m3u8  │ avc1.4D400B 1313k video only
  244      webm  360x480     30    │ ~  6.55MiB  753k dash  │ vp9          753k video only         DASH video, webm_dash

And then download each of the "audio only" entries from the table. In the case of the table for the video I chose:

  yt-dlp -f 233      -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"
  yt-dlp -f 234      -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"
  yt-dlp -f 139-dash -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"
  yt-dlp -f 139      -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"
  yt-dlp -f 140-dash -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"
  yt-dlp -f 251      -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"
  yt-dlp -f 140      -o "%(id)s.%(format_id)s.%(ext)s" "https://youtube.com/watch?v=a28_aXgrgXE"

Here I used the '-f' option to choose each of the 'audio only' formats available for the example video, and then I used the '-o' flag to specify a custom format string for the output files so that the file names include the format id making them unique from each other and corresponding to the entries in the original table.

This gives me files containing each of the audio formats that were available from YouTube.

  -rw-r--r--  1 user user 437246 Oct  9  2013 a28_aXgrgXE.139-dash.m4a
  -rw-r--r--  1 user user 437246 Oct  9  2013 a28_aXgrgXE.139.m4a
  -rw-r--r--  1 user user 713133 Oct  9  2013 a28_aXgrgXE.140-dash.m4a
  -rw-r--r--  1 user user 872935 Oct  9  2013 a28_aXgrgXE.140.m4a
  -rw-r--r--  1 user user 441481 Oct  9  2013 a28_aXgrgXE.233.mp4
  -rw-r--r--  1 user user 881428 Oct  9  2013 a28_aXgrgXE.234.mp4
  -rw-r--r--  1 user user 711273 Jul 22  2019 a28_aXgrgXE.251.webm

The timestamps of the files are set by yt-dlp to correspond to timestamps it got from YouTube.

It might be worth to be careful about downloading alternate format versions of too many videos. I could imagine that downloading alternate formats of too many videos from YouTube could trigger something on their side to make them think you are a bot or something. Of course that's just speculation and I don't know if YouTube actually does that. Hopefully doing it for a single video won't get your IP banned by YouTube.