Make all your songs the same volume
Ten songs evened out to one loudness, the quiet one boosted without distorting.
A real run on 13 November 2026, on a Linux computer, using Claude Code with Opus, on a Claude plan. The waiting is sped up; it took 1 min 50 s in real time.
The short answer.
Match loudness, not peaks. Audacity's Loudness Normalization (set to about −14 to −16 LUFS) evens songs out by how loud they sound; its plain Normalize only lines up the highest peaks. A quiet song with sharp peaks needs a limiter, or boosting it distorts. Stella (stella.sh) evened out a ten-song party mix to the same loudness, kept a backup of the originals, and turned down a quiet song's drum hits slightly so they wouldn't clip.
Stella does this on Windows, Mac and Linux.
What happened.
Ten songs evened out in about 1 minute 50 seconds. Checked afterwards against an answer key written before the run: all 10 within 2 LU of each other (they ended within 0.1 LU, from a 19.8 LU spread), at a normal listening level of −16 LUFS, no clipping, every song the same length with the same music, and titles and artists kept.
The ten songs are fictitious, synthesised for this run at deliberately different loudness. Our check script was fixed after the run to count files that kept their original dates, and re-scored; the answer key didn't change. The work is sped up 5×.
How Stella does it.
- Finds the Party mix folder in Music and measures each song's loudness: from very quiet to one far too loud
- Finds that the quietest song, Glasshouse, has sharp drum peaks that would distort if simply turned up
- Copies every original into Party mix (originals before loudness fix) before changing anything
- Brings the nine songs that needed it to one loudness, using a limiter on the quiet one, saves them at high quality and keeps titles and artists; leaves Saltwater Radio, already at the right level, alone
- Says which songs changed and how, and how to put the originals back
The prompt.
The songs in my Party mix folder in Music are all different volumes. Some are really quiet and one is way too loud. Make them all the same loudness.
This is the exact prompt from the recording. If your files are somewhere else, say where.
Do it yourself.
Audacity: Loudness Normalization
Open a song, select all, choose Effect, Volume and Compression, Loudness Normalization, and set perceived loudness to −14 or −16 LUFS. Export it, then repeat for each song, or record the steps as a macro (Tools, Macro Manager) and apply it to the folder.
Use a limiter on quiet songs with sharp peaks
If a boosted song crackles on loud hits, undo, then run Effect, Volume and Compression, Limiter (limit to −1 dB) before normalizing. That shaves the peaks so the rest can come up without clipping.
ffmpeg
ffmpeg -i in.mp3 -af loudnorm=I=-16:TP=-1.5:LRA=11 -c:a libmp3lame -q:a 0 out.mp3sets one song to −16 LUFS with a true-peak ceiling. Run it for each file; the two-pass form, measuring first, is more accurate.Keep a copy first
Normalizing rewrites the audio. Copy the folder before you start, so you can go back if a song sounds worse.
Or let the player do it
Apple Music's Sound Check and Spotify's volume normalisation even out playback without touching files, but only inside that app. For a USB stick, a party speaker or a DJ set, the files themselves need fixing.
Where it stops.
- The songs are rewritten. Stella kept the originals in a backup folder, which it said how to restore from; delete it once you're happy.
- The quietest, peakiest song had its loudest hits turned down a little to avoid distortion, so they may sound slightly less punchy.
- File names, tags and loudness measurements go to the AI model you use in Stella; the songs are processed on your computer.
- Recorded on Linux with ffmpeg already installed. On a Mac or Windows PC Stella would usually download ffmpeg first, which isn't shown.
Questions.
What's the difference between normalize and loudness normalize?
Peak normalizing makes each song's single loudest moment the same, which doesn't make them sound equally loud. Loudness normalizing matches how loud they sound overall (in LUFS), which is what you want for a playlist.
What loudness should I use?
Around −14 LUFS matches most streaming services; −16 LUFS leaves a little more headroom. The run used −16, and every song landed within 0.1 of it.
Is MP3Gain still a good option?
It works by changing a gain value inside MP3s without re-encoding, which is clever, but it's old, Windows-focused, and doesn't limit peaks. Loudness normalization with a limiter handles quiet, peaky songs better.
Can it fix song names and covers too?
Yes, in the same chat: see edit MP3 tags in bulk. To merge the mix into one file, see merge MP3 files.
Is Stella free for this?
Yes. Free with the ChatGPT or Claude plan you already have. Sign in with your ChatGPT plan, or use your Claude plan through Claude Code, and this task costs nothing extra. New accounts also get a small free trial of Stella’s own models.
Can I control what Stella changes on my computer?
Yes. Stella works on Windows, Mac and Linux with your own files and shows each step in the chat, so you can watch it and stop it at any point. Tell it what it may not do, as the prompts on these pages do (“don't delete anything without asking”). In the Downloads recording it deleted nothing until it was told to go ahead. It also refuses commands that would wipe a whole disk, your home folder or system folders.
Checked against these sources on 13 November 2026.