A music generation AI developed by Google DeepMind. Just describe the mood, genre, and instruments you want in text, and it produces a track that flows naturally from intro to chorus to bridge. It was first introduced in November 2023 alongside YouTube’s experimental “Dream Track” feature, and after several model generations it is now available across Google’s products as of 2026 — the Gemini app, Google AI Studio, Vertex AI, Google Vids, and Google Flow. Every generated track carries a SynthID watermark, so it can later be identified as AI-generated music.
Key Features
- Song generation from text prompts: Write something like “quiet lo-fi hip hop with rain sounds, 90 BPM” and you get a track. Tempo, dynamics, instrumentation, and vocal style can all be specified in words
- Vocals and multilingual singing: It generates not only instrumentals but songs with vocals, supporting singing in multiple languages across a wide range of genres
- From short clips to full-length tracks: There is a clip model for roughly 30-second pieces, and a Pro model that lets you specify song structure — verse, chorus, bridge — for tracks up to about three minutes
- Music from images: Upload a photo or illustration and generate a track that matches its atmosphere
- Real-time generation (Lyria RealTime): A separate model dedicated to interactive generation, where the music shifts continuously as you adjust the prompt
- SynthID watermarking: An imperceptible watermark is embedded in every track, allowing AI-generated audio to be detected
- 44.1 kHz stereo output: MP3 by default, with uncompressed WAV available on the Pro model
Pricing
Lyria is not sold as a standalone service. It is bundled into paid Gemini app plans and into Google Cloud’s API billing.
| Access route | Price | What’s included |
|---|---|---|
| Gemini app (Google AI Plus / Pro / Ultra) | Included in each plan’s monthly fee | Daily track limits vary by plan (higher tiers allow more). Longer tracks via the Pro model are available on upper tiers |
| Gemini API (Lyria 3 Pro Preview) | About $0.08 per song | For developers. MP3 / WAV output, prompt-driven song structure control |
| Vertex AI | Google Cloud usage-based billing | For enterprises, used within an existing Google Cloud contract |
Pricing reflects information as of August 2026. Plan names, monthly fees, and daily generation limits are subject to change, so check the official site and the Gemini and Google Cloud pricing pages for the latest details.
Pros & Cons
✅ Pros
- You can get a polished track by writing words alone, with no music theory or DAW knowledge required
- Available from within the Gemini app, so there is no new tool to sign up for or learn
- 44.1 kHz stereo output is good enough to use as-is for video background music and prototypes
- The SynthID watermark makes it possible to confirm afterwards that a track is AI-generated
- Via the API or Vertex AI, it can be embedded into your own service or production workflow
⚠️ Cons
- Track length is capped, so it is not suited to producing pieces running well beyond a few minutes
- Editing controls are limited — swapping out a single part or remixing means going back to a DAW
- Commercial use rights depend on the terms of the specific access route (Gemini app / API / Vertex AI), so check before using output in client work
- Daily generation limits vary by plan, and heavy trial-and-error can hit the ceiling quickly
- Features are spread across several Google products, making it hard to see at a glance what is available where
Comparison with Similar Services
| Criteria | Lyria | Suno | Udio | Stable Audio |
|---|---|---|---|---|
| Provider | Google DeepMind | Suno | Udio | Stability AI |
| Main entry point | Gemini app / API / Vertex AI | Dedicated web app | Dedicated web app | Web app / API |
| Songs with vocals | Yes | Yes | Yes | Mostly instrumental |
| Generation from images | Yes | No | No | No |
| Real-time generation | Yes (Lyria RealTime) | No | No | No |
| Strength | Integration with Google products | Ease and polish of song generation | Rich editing and remixing tools | Sound effects and loop material |
Who Is It For
- People who want to make their own background music for videos or slides instead of hunting through stock libraries
- Existing Gemini paid-plan subscribers who want to try music generation without another subscription
- Songwriting and lyric-writing learners who want to churn out demo tracks and idea sketches quickly
- Developers looking to embed music generation into their own app or content production pipeline
- Creators working with music that changes in real time, such as installations or live visuals
Summary
Lyria is Google DeepMind’s music generation AI, capable of producing anything from instrumentals to songs with vocals out of text or images. Its defining trait is that it is not a standalone service but is embedded across Google’s products — the Gemini app, the API, Vertex AI, Google Vids — which makes the barrier to entry especially low for anyone already using Google’s AI. On the other hand, track length and editing freedom have real limits, and fine-grained polishing still assumes DAW work. A good starting point is to generate a few tracks in the Gemini app and judge with your own ears whether they hold up for video background music or demo production.