Talk is cheap. Singing is free.

Voices that sing what you write and say what you type, made in your browser as you listen. No server.

It talks, too

It talks, too

Made right here, on your own computer. There is no server to run and no cloud bill, so the voice can ship inside your product and speak on your users' devices.

Press Speak. How fast it was made on your computer shows here.

The first press loads the engine and a voice, about 57 MB.

The editor

Write songs for up to sixteen singers

Put notes on a piano roll with a syllable under each, draw the vibrato or copy a trained singer's, then add singers and mix them. It sings each change as you make it.

  • Each singer in a tab, with a mixer and a master chain
  • MIDI files in and out, their lyrics with them
  • Parts and scores saved as plain text files
The editor's example: the first eight bars of Row, Row, Row Your Boat, a round for Finn, Cora and Ben.

Numbers

  • 12x faster than real time on one CPU core
  • 0.1 s to first sound, natively
  • 1.1 MB engine, with 56 MB of models before the first word
  • 0 third-party dependencies

Measured on one core of an Apple M3 Max, November 18, 2026. It runs in a browser, and natively on macOS and Linux from Python or the command line.

Get started

pip install lipservice
import lipservice as lip
lip.say("Hello from your own CPU.").save("hello.wav")
lip.sing("bpm 90\nTwinkle C4:1 C4:1\ntwinkle G4:1 G4:1\n", voice="cora").save("star.wav")

The first call fetches the models, 56 MB, and keeps them.

Read the docs

Free under the AGPL. Licensed for everything else.

Lip Service is free under the GNU AGPL. If you're shipping it in a closed product or service, a commercial license replaces the AGPL's terms with a simple contract.

See licensing