Gemini can now generate a 30-second approximation of what real music sounds like
Source: Engadget
New Lyria 3 Model
Google announced that its newly incorporated Lyria 3 model enables Gemini users to generate 30‑second music tracks from a prompt or remix an existing track. The model extends Gemini’s existing capabilities for text, image, and video generation and will be available in YouTube’s “Dream Track” feature for creating detailed backing tracks for Shorts.
Prompting and Control
- Example prompt from Google: “a comical R&B slow jam about a sock finding their match.”
- Users can specify granular elements such as tempo or drumming style.
- Gemini can also generate music from a photo or video, and the resulting tracks can be paired with album art created by Google’s Nano Banana image model.
Improvements Over Previous Models
- More realistic and musically complex tracks.
- Greater control over individual song components.
- Automatic lyric generation.
Outputs remain limited to 30‑second clips, but the promotional video suggests the possibility of longer clips or integration into other apps (e.g., Google Messages).
Watermarking and Detection
Songs generated with Lyria 3 are watermarked with Google’s SynthID, making it harder to pass them off as human‑created. Google introduced a SynthID Detector at Google I/O 2025 for identifying AI‑generated content.
Quality Assessment
- Instrumental sections are generally high‑quality.
- Lyrics can sound “corny” or “strange,” reflecting the current limits of the model’s lyric generation.
Availability
- Prompting tracks in Gemini is open starting today.
- Eligibility: users must be 18 years or older and speak English, Spanish, German, French, Hindi, Japanese, Korean, or Portuguese.
This article originally appeared on Engadget: Gemini can now generate a 30‑second approximation of what real music sounds like.