How to Create Songs with Lyria 3.5 in Gemini

Google DeepMind released Lyria 3.5 in September 2026 as its most advanced music generation model. It produces richer arrangements and more expressive vocals than earlier versions. You can use it right now inside the Gemini app on web or mobile without leaving the chat.

This guide shows exactly how to start a track, choose vocals or instrumental, pick length, and refine results with better prompts. All steps come from Google’s official launch post and product documentation.

What Lyria 3.5 adds

Lyria 3.5 improves on previous Lyria models with higher-fidelity audio and more natural vocal performance. You can select a genre or describe one in plain language. Templates help you jump straight into common uses such as background music or birthday songs. Track length is now flexible: short clips around one minute or longer pieces that run two to three minutes.

The model is available to every Gemini user globally on the web and in the Android and iOS apps. Higher Google AI plan tiers raise weekly generation limits, but the core feature works on the free tier.

How to open Create Music

1. Sign in to the Gemini app or open gemini.google.com in a browser.
2. Start a new chat or continue an existing one.
3. Look for the Create music option. On mobile, tap the + icon, then choose Create music. On web, open Tools or the attachment menu and select the same option.
4. The interface loads a prompt box with genre, style, and length controls.

If the button is missing, update the Gemini app or refresh the page. The feature rolled out globally after the 4 September 2026 announcement.

Step-by-step: generate a track from text

1. Choose Vocals or Instrumental.
2. Select a genre from the list or type your own description (for example, “upbeat indie folk with acoustic guitar”).
3. Pick short or full length.
4. Optionally choose a template such as background track or custom jingle.
5. Write a clear prompt that includes mood, instruments, and any story details. Example: “Create a calm instrumental track for a morning study session with soft piano and light strings.”
6. Send the prompt. Gemini generates the audio along with cover art created by Nano Banana.

Download the result as MP3 or as an MP4 that includes the cover. You can then ask Gemini to revise tempo, add a verse, or change the genre in a follow-up message.

Turn a photo or video into a soundtrack

Upload an image or short video before you send the prompt. Gemini analyzes the visual mood and composes a matching track. A sunset photo tends to produce warmer, slower arrangements. A busy city scene often leans toward energetic rhythms.

Add a short direction with the upload: “Instrumental cinematic score for this hiking video, building gradually.” The model returns a track whose style aligns with the visual content. This works for both short and longer lengths.

Tips for stronger results

• Specify instruments, tempo, and mood in one sentence. Vague prompts produce generic tracks.
• Mention whether you want lyrics or pure instrumental before generation starts.
• Use follow-up messages to refine: “Make the chorus more upbeat and add a brief guitar solo.”
• Check your Google AI plan limits if generation fails after several tracks in one day.
• For developer use, the same model is available through the Gemini API and Google AI Studio. See our related guide on Gemini API music generation for code examples.

Lyria 3.5 also appears in Google Flow Music and Google Vids, so tracks you create in Gemini can move into those tools for further editing.

Limits and availability

Free users receive a weekly token allowance that refreshes automatically. Google AI Plus, Pro, and Ultra subscribers receive higher limits. The feature requires a signed-in Google account and is restricted to users 18 and older in some regions.

Generated audio includes SynthID watermarks. Do not claim the tracks as fully human-created when you share them. Review Google’s usage policies for commercial distribution.

Conclusion

Lyria 3.5 turns the Gemini app into a practical music tool for quick jingles, video soundtracks, or personal experiments. Start with a clear prompt, choose vocals or instrumental, and iterate with follow-ups. The same workflow works on web and mobile.

For more Gemini productivity ideas, check our guide on creating skills for repetitive tasks.

Sources

• Create your best tracks yet with Lyria 3.5 in Gemini
• Lyria model page
• Gemini API music generation docs