Mastodon Feed: Post

Reblogged by jsonstein@masto.deoan.org ("Jeff Sonstein"):

nina_kali_nina@tech.lgbt ("Nina Kalinina") wrote:

"LOL I EXPORTED A SONG AS JPEG" aka
Experimental answer to "should/could you use JPEG instead of MP3 for music compression?"

Boring answer: "shouldn't, but could"

Fun answer: let's try!

There were two things I wanted to test:

1. Audio treated as RGB vs audio treated as greyscale JPEG. JPEG _could_ compress color component of the image more than luma, so greyscale JPEG should give better audio quality.
2. How well JPEG will fare compared to MP3.

RAW audio was 485 KB.
MP3 "insane" was 354KB.
MP3 "VBR low" was 103KB.
MP3 "fixed 8kbps" was 45KB and still OK to listen to.

JPG RGB 90% was 117KB and the quality was between low VBR and fixed. It got noise though.

Grey 90% at 174KB was better than VBR low, but got some noise too.

Grey Potato (0%) at 11KB was horror.
Grey Less Potato (10%) at 30KB resembles very poor radio reception.

This is Grey 90% JPEG converted back to WAV:
https://drive.google.com/file/d/1gtxBxSbvQt4z0Djhw_7vUo5xKdqcGSi7

This is "less than 8kbps MP3" Grey Less Potato: https://drive.google.com/file/d/1Yco3FkrwE6C_5MiOs1sFGkL9u3n1-Js5

All files: https://drive.google.com/drive/folders/1hXDx0OKqLY2XZ2qlpCFjdPO9byM8ombm

Attachments:

• "Noise" - it's a short song encoded in greyscale JPEG (remote)
• The same song, encoded in RGB JPEG. Also noise with some patters (remote)