[LAU] Transcoding from lossy to lossy with minimal loss...

Hello, I have a server with limited storage that I want to run a private radio station from, a randomized mix of my complete music collection. Locally I have about 80G of music in all sorts of formats, codecs and bitrates. This is way too large for the server's storage, I can use half of that at best. Additionally I don't want the stream to have too much bandwidth so it will work even over flaky (mobile) network connections. My thought is to transcode all of it to the same reduced format, then upload.  That way the music server could just push it out without transcoding again (and I could still listen to separate tracks remotely). The Big Question: Which format should I choose? I found these 2 articles that seem to have an answer: https://wiki.hydrogenaud.io/index.php?title=Transcoding#Lossy-to-lossy_tran… https://trac.ffmpeg.org/wiki/Encode/HighQualityAudio Combined, it sounds to me like I really should use either FDK AAC or Opus* at less than 100kbps (I listen to 64k AAC music streams that are OK imo). What do you think? Is this even the right approach to solve the problem? TIA! FWIW, here's a breakdown of my music's codecs/bitrates: vorbis: 97 files (974M), average bitrate 332kbps (from 67 to 452) wmav2: 10 files (31M), average bitrate 131kbps (from 129 to 133) flac: 1216 files (45505M), average bitrate 1191kbps (from 330 to 5170) opus: 173 files (1024M), average bitrate 129kbps (from 76 to 177) mp3: 1975 files (32823M), average bitrate 197kbps (from 96 to 420) aac: 308 files (1592M), average bitrate 152kbps (from 64 to 334) alac: 1 files (621M), average bitrate 768kbps (from 768 to 768) * personally, I always had the feeling that opus (used a lot by youtube) isn't so good with noisy, grungy, fuzzy, guitarry music