Same almost. I have an ~800gb main library of mostly lossless files that I squash to around 150gb by transcoding to 196k or something opus that i put locally on my phone. I also strip embedded cover art which can save a stupid amount of space sometimes; relying on folder hierarchy with cover.jpg/png files. (Bitrate is pretty overkill for me so I may drop it to 128-160…)
I haven’t had the time to manage the tags properly on my reference library*, but my folder hierarchy encodes artist/album/title with optional years and track numbers. I wrote a linter script to check the structure, that every folder has a cover art image, and to warn about lossy formats not in directories suffixed with [lossy] (purely for documentation purposes; not used in script logic).
My transcode script generates tags from the folder and filenames, only copying genre tags if they exist and stripping everything else. Lossless files are transcoded while structure, art, and lossy files are copied. Then that result is synced to my mobile devices. So whenever I add music my workflow is to just name file folders properly and download or extract art then I just lint, transcode, and then resync.
*(Tags of my reference library don’t matter so much to me, but the squashed lib needs consistent tags for mobile apps for behave as I intend)
I hadn’t bothered because it was mostly LLM slop I edited until it worked, so it’s specific and not robust. Just had to hack something together the night before I had to leave on a trip… Haven’t had time to tag my music properly, so I’m still using it. The magic sauce is just ffmpeg and the mutagen Python lib.
https://pastebin.com/eam6Ydz1
https://pastebin.com/kL6pu5Wt