Lyria 3.5 (Music)
Google DeepMind's Lyria 3.5 composes a full song, with sung lyrics or instrumental, from a text prompt and an optional picture that sets the mood.
Lyria 3.5 is Google DeepMind's music model. From a text description it composes a complete, structured song up to a few minutes long, with vocals and lyrics or as an instrumental, in English, German, Spanish, French, Hindi, Japanese, Korean or Portuguese. It placed sixth on both Artificial Analysis music arenas, vocals and instrumental, in November 2026.
Put the genre, mood, instruments, tempo, vocals and structure into the prompt, up to 5000 characters. Ask for a length with a hint such as 'a 2-minute track' or with timed sections such as [0:00-0:30] Intro, and put the words to sing in quotes. An image, if you add one, sets the mood and theme of the music.
Inputs: prompt (required) and image (optional, a link or an upload). Outputs: audio (the song, an MP3 file) and lyrics, the words of the song with their timing, when it has any. One song per request.
Price: $0.10 per song. It makes music only: it does not take a reference song, a melody or a negative prompt, and it does not make speech or sound effects.
- Music Generation
Price: $0.10 per run. One run with default settings costs $0.10.
Run Lyria 3.5 (Music) via API
curl -X POST https://api.flymy.ai/api/v1/flymyai/lyria-3_5-music/predict \
-H "X-API-KEY: $FLYMYAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"prompt":"A 30-second upbeat acoustic folk instrumental with guitar, mandolin and hand claps"}'API documentation · Create a free account · Pricing