Drop your .mid onto Audial's Text to Vocal piano roll (or into the free text2vox plugin), type the lyrics, add 5 to 10 seconds of the voice you want, and Audial sings the melody note by note, returning a WAV of the vocal and the MIDI it sang.
$10 a month or $100 a year. Unlimited runs, cancel any time. 2 free runs to try it first.
MIDI to vocals is the case where the melody is already written. Export the lead line from your DAW as a .mid and drop it onto the Text to Vocal piano roll: the notes land in the roll with their timing, ready to sing. Add 5 to 10 seconds of the voice you want as the reference, type the words, and render. You get a WAV of the sung vocal and the MIDI it followed.
For exact placement, type the lyrics per note: one token per note, separated by |, with - to hold a syllable through the next note, so one word can span several notes, as in a|ma|zing|grace|-|how|sweet|the|sound. On the website you can also type a syllable straight under each note in the roll. If you would rather not count notes, auto mode spreads plain text across the melody for you.
In your DAW, text2vox is the same tool as a free VST3 and standalone plugin for macOS and Windows, with the same piano roll and .mid import. From code, send the file as text2vox.midi to /functions/run/text2vox, or pass midi_file to audial.text2vox in Python.
What you get
- Drop a .mid on the roll; every track in the file merges into one sung line
- Lyrics one token per note with holds, typed under the notes, or spread automatically
- A 5 to 10 second reference clip sets the voice
- Melodies up to 2 minutes per render
- Returns a WAV of the vocal and the MIDI that was sung
- Free text2vox plugin for your DAW, plus the REST API, the Python SDK and MCP
Hear it
The reference voice came from the Audial music model. The melody was a 9-note MIDI file and the lyrics were typed.
Audial vs doing it by hand
| Task | Audial | By hand |
|---|---|---|
| Hearing the melody sung | Audial: Rendered from your MIDI, your words and a short voice clip | By hand: Sing it yourself, or book a singer |
| Changing the melody | Audial: Edit the notes in the roll, or export a new .mid, and render again | By hand: Learn the new part and sing it again |
| Lyric placement | Audial: Exactly one syllable per note when you type it per note | By hand: A singer phrases it their own way, for better and for worse |
| Performance | Audial: A synthetic vocal: right for demos, guide vocals and ideas | By hand: A good singer brings phrasing and feeling no render matches; for a release, that still wins |
| Cost | Audial: 2 free runs, then $10 a month with unlimited renders | By hand: A singer's fee and studio time, or your own time and voice |
Use it from code
# One token per MIDI note; "-" holds the syllable before it through that note.
curl -X POST "https://api.audialmusic.ai/api/functions/run/text2vox" \
-H "X-API-Key: your_api_key" -H "X-User-ID: your_user_id" \
-F "userId=your_user_id" \
-F "original=@/path/to/voice.wav" \
-F "text2vox.midi=@/path/to/melody.mid" \
-F "text2vox.lyrics=a|ma|zing|grace|-|how|sweet|the|sound" \
-F "text2vox.lyricsMode=per_note"import audial
audial.config.set_api_key("your_api_key")
audial.config.set_user_id("your_user_id")
result = audial.text2vox(
reference_file="voice_clip.wav",
midi_file="melody.mid",
lyrics="a|ma|zing|grace|-|how|sweet|the|sound", # one token per note
lyrics_mode="per_note",
)
print(result["files"]["folder"]) # the sung WAV and the MIDI it followed# In Claude, Cursor or any MCP client with audial-mcp installed:
"Sing 'amazing grace how sweet the sound' over ~/Music/lead.mid using ~/Music/voice.wav as the voice."Frequently asked questions
Open Text to Vocal, drop the .mid onto the piano roll, add a 5 to 10 second clip of the voice you want, type the lyrics and press Render. The MIDI must be one note at a time: a single melody line, not chords or harmony. You get a WAV of the sung vocal and the MIDI it followed. The free text2vox plugin does the same inside your DAW.
Type one token per note, separated by |. A - (or + or an empty token) holds the syllable before it through that note, so a word can span several notes. The first token must be a syllable, and the count must match the notes; there is no rest token, so leave silent notes out of the MIDI. Or let auto mode spread plain text across the melody.
It sings one line. Every track in an imported file merges into one voice, so export the lead melody on its own. For a harmony, render each line as its own vocal and stack them in your DAW.
Up to 2 minutes per render. For a longer part, render it in sections and join them in your DAW.
The voice in the reference clip you upload. 5 to 10 seconds of that voice on its own works best. Only use a voice you have the right to use: a reference clip is a likeness.
Every account gets 2 free runs. After that, rendering needs a subscription: $10 a month or $100 a year, with unlimited runs. The text2vox plugin is a free download and uses the same subscription.