Google Lyria 3 Generates Songs from Text & Images

Google has announced one of the biggest updates in AI-based music production, integrating the Lyria 3 model into the Gemini app. With this feature, users can now create original 30-second songs simply by writing an idea or uploading an image. This feature stands out with its multimodal structure that evaluates both text and visual inputs, its ability to automatically generate lyrics, and offering detailed control over tempo, genre, and vocals.

Key Features of Lyria 3

Text to Music: Turn Ideas into Songs

In the Gemini app, simply writing a command like “a peaceful, nostalgic lo-fi track” or “a funny R&B slow jam” is enough for Lyria 3 to generate a vocal or instrumental piece. The model produces a high-quality, approximately 30-second clip. Lyria 3 has been enhanced to create more realistic and complex compositions compared to earlier versions.

Photo/Video to Music: Multimodal Flow

Not only text, but also photos or short videos can be uploaded. The AI analyzes the context and atmosphere of the content to produce a piece with matching lyrics and melody. This creates a seamless visual-audio storytelling experience.

Automatic Lyrics Generation

Lyria 3 takes over the task of writing lyrics. Based on the theme you provide, it generates the lyrics and synchronizes them with the melody it creates.

More Control: BPM, Genre, and Vocals

Users can guide the result by specifying details such as BPM (tempo), music genre, and vocal characteristics. This provides content creators with a wide range of expressive possibilities, from short jingles to emotional ballads.

Automatic Cover Art and Easy Sharing

For every track created, Gemini generates a unique cover art, making it easy to instantly share the content with a visual component on social media and video platforms.

Security and Content Transparency: SynthID Watermark

Google requires all music produced with Lyria 3 to include a SynthID watermark, an invisible marker that can later verify that the file was AI-generated. This approach is crucial for ensuring the verifiability of AI-generated content and addressing copyright concerns.

Use Cases: Ideas for Marketing, Events, and Content Creation

• Social Videos/Shorts Background Music: Quickly create short, copyright-free jingles (30-second tracks) for YouTube’s Dream Track and Shorts experiences.
• Event Intros & Presentations: Produce a short intro that reflects the theme or brand tone of an event, and share it instantly with the cover art.
• Personal Storytelling: Generate a nostalgic track from a photo to add a musical frame to memories. The model is inspired by the visual context.

Access, Languages, and Age Limit

According to Google, the Lyria 3 feature is currently available as a beta in the Gemini app for users aged 18 and above. The feature supports multiple languages, with English being the primary language highlighted in the announcement. The Workspace update notes that the 30-second custom music creation feature will be gradually rolled out starting February 18, 2026, and can be controlled from the admin settings.

How to Use It? (Step-by-Step)

1. Open the Gemini app and select “Create music” from the tool menu.
2. Write a text prompt (e.g., “Energetic, 102 BPM, female vocals, 2000s pop vibe”). Alternatively, upload a photo/video; Lyria 3 will consider the visual context.
3. Specify parameters like style, BPM, and vocals; the model will apply these instructions.
4. Listen to the approximately 30-second generated track; download and share it along with the cover art.
5. The file will include the SynthID watermark, which can be used to verify that the content is AI-generated.