Lyria 3

Lyria 3 is Google’s music generation model, available through the Gemini API. It generates high-quality, 48kHz stereo audio from text prompts or images.

This is the Clip variant — it always generates a 30-second MP3 clip. For full-length songs (up to ~3 minutes), use Lyria 3 Pro.

What you can do

• Generate music in any genre: pop, jazz, electronic, orchestral, lo-fi, and more
• Specify instruments, tempo (BPM), key, mood, and style
• Provide up to 10 images as inspiration — the model composes music based on visual content
• Generate instrumental tracks or tracks with vocals and lyrics
• Prompt in different languages to get lyrics in that language

Tips

• Be specific: mention genre, instruments, BPM, key, and mood for best results
• Use “Instrumental only, no vocals” if you don’t want vocals
• Include structure tags like [Verse], [Chorus], [Bridge] to guide composition
• The more detail you provide, the better the output

Output

• 30-second MP3 audio clip
• 48kHz stereo
• All output includes SynthID watermarking

Limitations

• Always generates 30-second clips (use Lyria 3 Pro for longer tracks)
• Single-turn generation only — no iterative editing
• Some prompts may be blocked by safety filters
• Results vary between calls, even with the same prompt