Lyria 3 is Google’s most advanced music generation model. It can create songs up to 3 minutes in length, complete with instruments, vocals, and lyrics, from both text and image prompts.
This model is notable for offering professional-grade production and audio, and for offering wider commercial use rights than any other AI-music model
Lyria Standard vs Lyria Pro
- Lyria 3 Standard can output songs with a duration of 30 seconds. Usually, these songs don't have a proper structure to them, starting and ending abruptly
- Lyria 3 Pro can output songs between 30 seconds and ~3 minutes (Full Song). These songs are structured and have a proper beginning and end.
Key Features
- High-Fidelity Music Generation:
- Generates high-quality music and audio, including complex tracks with diverse instrumentation and vocal elements.
- Instrumentals and vocals
- Create songs with singing or without. Supply your own lyrics or have the AI model auto-generate lyrics automatically
- Free-text Prompting
- This model is reason-capable: prompt it in any level of detail you want in natural language, including lyrics, genre, song length, bpm, subject, function (underscore, trailer, commercial…), instrumentation, inspiration, time era, timestamps for climax/drop, etc
- Create full-length songs
- Create songs from 30 seconds up to 3 minutes
Artlist Features
These features are only available on Artlist:
- Artlist Sound:
- A system of prompt enhancements that improves the quality and style of the model, taking it from stock to style.
- Settings:
- Set Genre, Mood, Theme (use case), and Tempo from drop-down menus. Users are not limited to the dropdowns and can also manually prompt these via text.
- Auto Prompt and Auto Lyrics:
- Auto-generate the prompt or the lyrics based on existing user inputs.
Limitations
Can’t prompt with famous artists or song names:
- Using artist names or song names in the prompt will block the generation. Instead of “Song in the style of Taylor Swift”, write “A country-pop crossover with acoustic guitar and a catchy melodic hook with storytelling lyrics, bright, airy vocals.”
No Audio inputs:
- No ability to upload audio or other sonic references
Technical Capabilities
- Modalities: Text-to-Audio, Image-to-Audio
- Vocals: Supported
- Lyrics: auto/custom/none
- Output Length Lyria Standard: 30s
- Output Length Lyria Pro: 30s up to Full song (~3 minutes)
- Output formats: wav 48kHz 16bit
- Language Support: All, multilingual
Prompt tips:
- Use the following prompt format: A mood genre song about subject.
- For example: An epic rock ballad about dogs
- For example: A soft indie song about love
- Describe specific genres, eras, and style: 80’s hair metal, 2000’s singer-songwriter ballad, 1860’s Opera, mellow lo-fi.
- Be more detailed to get more accurate results: instead of “rock song”, prompt: “early 2000’s punk-rock band, lead male vocal singer with gritty voice, heavy distorted guitars, loud drum fills. The Chorus should be catchy and anthemic.”
Prompt adherence
Excellent adherence:
- General genre
- Vocalist gender, style, and mood
- Instrumentation
- Specific BPM
- Energy
- Song subject and theme (for lyrics)