Google Lyria ids and prices next to the Sume Music Router ids
Google lists Lyria 3.5 at $0.08 a song, a 30-second clip model at $0.04, and Lyria 3 Pro Preview at $0.08. Which of them Sume routes, and at what flat price.
Google's Gemini API lists three Lyria entries as of 2026-10-04: lyria-3.5 at $0.08 per song, lyria-3-clip-preview at $0.04 per 30-second clip, and lyria-3-pro-preview at $0.08 per full song, all paid-tier only. Sume's Music Router routes sume/music-auto (Lyria 3.5 today), lyria-3.5 and lyria-3-pro, and bills a fixed $0.125 per generation regardless of engine.
What Google lists
The Gemini API pricing page gives per-song prices and says the free tier is not available for these models. The music generation guide says lyria-3-clip-preview always makes 30-second clips, while lyria-3.5 makes full-length songs of a couple of minutes, controllable by prompt. Both are 44.1 kHz stereo, default to MP3, and carry a SynthID watermark.
| Google model id | Output | Price on the Gemini API | Free tier |
|---|---|---|---|
| lyria-3.5 | Full song, about a couple of minutes | $0.08 per song | Not available |
| lyria-3-clip-preview | Always a 30-second clip | $0.04 per song | Not available |
| lyria-3-pro-preview | Full song | $0.08 per song | Not available |
What Sume routes
The Music Router docs list routable ids sume/music-auto (the default), lyria-3.5 and lyria-3-pro, and say an unknown id fails with 400 model_not_found and a catalog_url. The docs do not list a 30-second clip model, so a short cue on Sume is a prompt that says so ("a 30-second track"), billed at the full per-generation price. GET /v1/music-router/models returns the live catalog, with the provider list price per model for reference.
| Sume id | Behaviour | Price per generation |
|---|---|---|
| sume/music-auto | Sume picks the engine; Lyria 3.5 today | $0.125 fixed |
| lyria-3.5 | Pinned pass-through | $0.125 fixed |
| lyria-3-pro | Pinned pass-through | $0.125 fixed |
Reading the gap
On a per-track basis Google's list price of $0.08 is $0.045 lower than Sume's $0.125, and for a 30-second cue Google's clip model lists $0.04. What Sume adds is the surrounding surface: one API key, async jobs with webhooks, hosted artifacts on media.sume.com, and Timeline 1.0 to put the track under video. If you call Google directly you build those parts yourself.
- Short cues: a 30-second prompt on Sume costs the same as a three-minute one.
- Reproducibility: pin
lyria-3.5instead ofsume/music-autoif the engine matters to your log. - Audit:
job.request.routed_modelnames the engine that ran.
Check before you budget
Google's preview models can change or retire, and Sume's catalog is the thing to query rather than a blog table. Treat every figure here as dated 2026-10-04. For polling and results, see Jobs and results.
Sources
Related posts
More in Models
- Google media model ids on Oct 4: Omni, Veo 3.1, Lyria 3.5 vs Sume ids
Which Google model id to call on Oct 4, 2026 and the matching Sume id: Omni Flash 1.1, Veo 3.1 previews, Lyria 3.5 and Lyria previews, with status and dates.
- GPT Image aspect ratios: Firefly's 10 vs Sume's 17
Firefly's September 2026 update gives GPT Image 2 ten aspect ratios instead of three. Sume's image API documents 17 ratios plus auto, and custom pixels.
- How long can a Seedance 2.5 product clip be on Sume?
On Sume, seedance-2.5 accepts 4 to 30 seconds at 480p, 720p or 1080p. How to pick a length for a product clip and poll the job until it completes.
- Hume Octave 2 covers 11 languages: is yours one?
Hume Octave 2 is reported to cover 11 languages. Before you pick a voice engine, check your language, then verify what Sume's tts_create recorded for the job.
Written by Sume