Audo is a music studio for Mac, for people who start with a voice. Move the notes that landed wrong inside your own recording, lay your lyrics on the melody, and build the track around it. The notes you leave alone stay exactly as you sang them.
Try it: drag the red note up. This is a sketch in your browser where a synth stands in for the voice. In Audo, the recording itself is re-pitched, one note at a time, with the voice's character kept.
Audo transcribes a draft from your vocal on your Mac, then you write over it. "Time to the vocal" lays each syllable on a sung note: a word can hold several notes or share one. It's the same take as above. Type your own line and time it, and the synced lyrics follow.
Two bars in F minor, the same notes as the chorus above. Turn the slides and rolls on and off while it plays, or click the grid to change the drums. Slides glide the 808 between notes that are close enough to slide, and add an octave-up pickup into the downbeat. Rolls go on a track of their own and skip any hit that would flam with your hats.
Audo keeps a log next to every project. Your edits and your writing are marked as yours. Finding notes, transcribing, timing and the one-click tools are marked as Audo's. Each entry carries the hash of the one before it, so changing an entry later shows. Try it below.
The log uses the app's format: SHA-256 over the previous entry's hash and the entry itself. Rewriting every hash after a change would get past this check, so Audo also keeps the newest hash on your Mac, and a rewritten log no longer ends where Audo left it. This demo's log stays in your browser and is gone when you reload.
The music is a demo project, written and rendered in Audo's own engine.
Move a sung note, transpose a stretch of a bass line, or slice a drum take at its hits. Each moved note is re-pitched on its own, and the voice keeps its character.
A draft transcribed from your vocal on your Mac, then your own words, each syllable timed to the notes you sang. Export LRC or SRT.
A wavetable synth, a drum machine and a multi-sample sampler, plus the Audio Unit instruments and effects you already own.
EQ with mid/side, compressor, delay, drive, limiter, stereo width and reverb, all automatable.
One click adds slides between bass notes and rolls into new sections. Running it again changes nothing, and muting the rolls track removes them.
A tamper-evident log tells your edits apart from the app's automatic steps, and every export is tagged.
Name a version, play any two against each other, and roll back anytime.
Mastering, loudness meters, and every version of your song (instrumental, clean edit, short cuts) in one click.
Your plugins' own interfaces, each running separately so one crash never stops the song.
Ask in plain words, and it edits through the same undoable steps you use, with every change in the log. An MCP server lets you do the same from a terminal.
Audo is built by one person who writes lyrics but had never been able to make the music under them, working with Claude Code. Several sessions work on the same repository at once, and review agents check each bigger change before it lands. The first commit was on September 29. By October 1 there were 45 commits and about 55,000 lines of Rust, TypeScript and Python.
Some of what that looked like: the first way of re-pitching a sung note buzzed on a vocal with heavy reverb, so each moved note is now re-pitched on its own with Signalsmith Stretch, its formants kept, and blended in over 20 ms. A review caught the app logging its own note-finding as the user's edit, which would have defeated the point of the log. And a test checks that the audio engine, written in Rust, never allocates memory or waits on a lock while it plays.
Builds go out a few at a time. Leave your email and you'll hear from me once, when yours is ready.