Lyria 3 API: Availability, Access Paths, and Pricing

Searching for a "Lyria 3 API" spec usually ends in confusion: Google rolled out Lyria 3 in the Gemini app, then Lyria 3 Pro in Vertex AI, then Lyria 3.5 in Flow Music — each with different access rules and price points. If you are a developer trying to figure out what is callable today, what it costs, and how to wire it up, this guide answers those three questions using Google's own documentation, announcements, and pricing pages. (Part of the Lyria3 Pro resource library.)

TL;DR

• What it is: Lyria 3 is Google DeepMind's music generation family, exposed to developers in the Gemini API as two preview models — lyria-3-clip-preview (30-second clips) and lyria-3-pro-preview (full-length songs).
• Where to access it: Gemini API (REST or Google Gen AI SDKs), Google AI Studio, and Vertex AI in public preview. Consumers also hit the same models via the Gemini app, YouTube's Dream Track, Google Vids, and ProducerAI.
• Official cost: $0.04 per 30-second clip (Lyria 3), $0.08 per full song (Lyria 3 Pro). Lyria 2 remains at $0.06 per generation.
• Output: 44.1 kHz stereo audio in MP3 (WAV on Pro), every track watermarked with SynthID.

What is the Lyria 3 API?

The Lyria 3 API is Google's programmatic interface to its Lyria 3 music generation models, served through the Gemini API (the generativelanguage.googleapis.com endpoint). Per Google's official docs: "Lyria 3 is Google's family of music generation models, available through the Gemini API. With Lyria 3, you can generate high-quality, 44.1 kHz stereo audio from text prompts or from images."

The family includes two models — picking the right one is your first decision:

Both models accept text prompts and up to 10 images, return lyrics and song structure alongside the audio, generate vocals in the language of your prompt, support instrumental tracks, and accept section tags or timestamps for structure control.

Two limitations worth knowing before you architect around them: generation is single-turn (no iterative editing via follow-up prompts), and prompts pass through safety filters — artist-voice requests or copyrighted lyrics are blocked.

Lyria 3 API availability: who gets it, and when

Lyria 3 has been rolling out in waves since February 2026, per Google's official announcements:

• February 18, 2026 — Lyria 3 launches in beta in the Gemini app, generating 30-second tracks from text or images (cover art via Nano Banana), and arrives on YouTube's Dream Track in the U.S., expanding internationally.
• March 25, 2026 — Lyria 3 Pro is introduced: tracks up to 3 minutes with structural awareness, available in Vertex AI (public preview), Google AI Studio, the Gemini API (alongside Lyria RealTime), Google Vids, the Gemini app (paid subscribers), and ProducerAI.
• July 29, 2026 — Lyria 3.5 is announced with improved musicality, lyrics, and vocal quality, debuting in Google Flow Music.

The practical takeaway for developers: the API surface today exposes the Lyria 3 Clip and Lyria 3 Pro preview models. Lyria 3.5 is a consumer-first release (Google Flow Music); as of this writing it is not listed among the Gemini API music generation models.

Developer vs. consumer access at a glance

How to access the Lyria 3 API: three paths, from quickest to most scalable

Path 1 — Google AI Studio. The fastest way to evaluate Lyria 3 without writing code: open AI Studio, pick a Lyria model, and prompt it — a good way to test prompts before spending API quota.

Path 2 — Gemini API with the Interactions API. Get an API key from AI Studio and call interactions.create. A minimal REST request:

curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "Content-Type: application/json" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -d '{
    "model": "lyria-3-clip-preview",
    "input": "A short instrumental acoustic guitar piece."
  }'

The same call works with the Google Gen AI SDK in Python or JavaScript (client.interactions.create(model="lyria-3-pro-preview", input="An epic cinematic orchestral piece...")). The response contains base64-encoded audio plus generated lyrics, exposed via output_audio and output_text.

Path 3 — Vertex AI. For businesses that need on-demand audio at scale, Lyria 3 Pro is in public preview on Vertex AI — Google positions it for generating bespoke game soundtracks or integrating into creative platforms.

Prompting rules of thumb that improve results: iterate on the faster Clip model before committing to a full-length Pro generation; be specific — genre, instruments, BPM, key, and mood all move output quality; specify duration in the prompt (e.g., "create a 2-minute song") or with timestamps.

Lyria 3 API pricing: the official numbers

Google's pricing page is the only authoritative source here, and it lists Lyria prices per generated output ("count"):

A quick cost sanity check (rule of thumb): 1,000 clips runs $40; 1,000 full songs runs $80. Google Cloud also states that you are charged only for requests that return a 200 response — failed or blocked requests aren't billed. Because both API models carry "preview" labels, treat these prices as current-but-not-permanent, and re-check the official pricing page before you commit to a budget.

Lyria 3 vs. Lyria 3.5: what to know before you build

Here is the one decision most coverage gets wrong: Lyria 3.5 is the newest model in the family (announced July 29, 2026), but it currently lives in Google Flow Music — Google's consumer music tool — not in the Gemini API. If your goal is an API integration today, you build on Lyria 3 Clip and Pro previews; if you want the latest vocal and musicality quality inside a consumer product, Lyria 3.5 is the release to watch. For a full breakdown of what changed between generations, read our explainer: What Is Lyria 3.5?.

FAQ

Is the Lyria 3 API free? No. Lyria generations are billed per output: $0.04 per 30-second clip and $0.08 per full song, per Google's official pricing page.

Does the Lyria 3 API generate vocals and lyrics? Yes. Lyrics are generated automatically from your prompt, or you can supply your own with [Verse], [Chorus], and [Bridge] tags. Lyrics follow the language of your prompt.

What's the difference between Lyria 3 Clip and Lyria 3 Pro? Clip always produces a 30-second MP3 — ideal for loops and previews. Pro produces full-length songs (a couple of minutes, prompt-controllable) with verse-chorus-bridge structure, and can return WAV instead of MP3.

Can I generate music from images through the API? Yes — both models accept up to 10 images alongside the text prompt.

Are Lyria 3 API outputs watermarked? Yes. All generated audio carries an imperceptible SynthID watermark for identifying AI-generated content.

Can I use Lyria 3 API output commercially? The models are trained on material Google and YouTube have rights to use, and outputs are designed not to mimic specific artists. Confirm your specific use case against Google's Gemini API terms and usage policies.

Sources

• Generate music with Lyria 3 — Gemini API docs (Google AI for Developers) — model IDs, 44.1 kHz stereo output, API examples, limitations. (Reviewed via archived capture dated July 26, 2026.)
• Lyria section of Google Cloud's generative AI pricing page — official prices: Lyria 3 Pro $0.08, Lyria 3 $0.04, Lyria 2 $0.06 per generation.
• A new way to express yourself: Gemini can now create music (blog.google, Feb 18, 2026) — Lyria 3 beta launch, Gemini app availability, SynthID.
• Lyria 3 Pro: Create longer tracks in more Google products (blog.google, Mar 25, 2026) — Lyria 3 Pro capabilities and availability across Vertex AI, AI Studio, and the Gemini API.
• Lyria — Google DeepMind model page — model family overview, Lyria 3.5 announcement, Flow Music availability.

Note on methodology: every number and date above is traced to a Google-owned source — developer docs, official announcements, or the Google Cloud pricing page. Lyria 3 API preview pricing may change; verify against the official pricing page before budgeting.