Can an AI assistant make music or songs?
Short answer
Gemini can: with Lyria 3.5 it makes songs with vocals and instruments from a text prompt or a photo, up to three minutes long, which you can download or share. You must be 18 or older, and Google says limits apply without giving numbers. We found no music feature for Claude: Anthropic's help pages list documents, diagrams, websites, charts and code, not audio. We couldn't open OpenAI's help pages, so ChatGPT isn't covered here yet.
Which products can do it?
- Google GeminiYes
Gemini makes songs with vocals and instruments from text or photos using Lyria 3.5, with tracks up to three minutes long that you can download or share.
Good to know: 18+, in all countries where the Gemini app is offered; limits apply. Every track carries Google's invisible SynthID watermark. In February 2026, tracks were still 30 seconds.
Why it matters
People want a custom song for a birthday, a video or a joke without an instrument, a studio or music software.
Not proven yet
How many tracks each plan allows; Google only says "limits apply". How the music sounds compared with dedicated music apps; no independent test in our sources.
How we know (2 sources)
Google's prompting tips for Lyria 3 in the Gemini app (read Oct 31, 2026): at the time, tracks were 30 seconds long; you can ask for vocals and pick a male or female singer, write your own lyrics by starting lines with "Lyrics:", or let the model write lyrics from a theme. Tracks download as MP3 or MP4. Since replaced by the longer Lyria 3.5 tracks on Google's overview page.
Read the sourceGoogle's music generation overview page (read Oct 31, 2026, English version): Gemini makes songs with vocals and instruments from text or photos using Lyria 3.5, which Google calls its best-sounding music model, with tracks up to three minutes long and a template gallery to start from. Tracks can be downloaded or shared. Available in all countries where the Gemini app is offered, for ages 18+; "limits apply", with no numbers and no split by plan. Every track made in the Gemini app carries SynthID, Google's invisible watermark, and you can upload an audio file to ask Gemini whether Google AI made it.
Read the source
Status timeline
No status has changed since we started tracking. Every entry is dated and backed by the sources on this page.
- First recorded: In storesGoogle Gemini
Gemini makes songs with vocals and instruments from text or photos using Lyria 3.5, with tracks up to three minutes long that you can download or share.