Applied artificial intelligence, neural systems, and smart agents.
Fusing generative AI, multi-agent frameworks, and high-throughput data processing to automate business workflows.
Lyria 3 Pro, the music model deployed on the Google API
Lyria 3 Pro is Google music model deployed on the Gemini API, and each full song is processed at a rate of eight cents — not by token, not by monthly subscription. It holds fifth position on the instrumental leaderboard and eighth on the vocals leaderboard, meaning its output performs more strongly under picture than out front.
- Eight-cent-per-song rate
- SynthID watermark
- No free layer
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. Change log
The technical specification, per Google official documentation
The model id is lyria-3-pro-preview. Output length is steered by the prompt, and the official documentation records a phrase rather than a precise figure — a couple of minutes; the Clip model in the same family produces exactly thirty seconds and is billed at a four-cent rate. Output is 44.1 kHz stereo, mp3 by default, with a WAV format available on the Pro model. Lyrics generation runs in the same pipeline and is returned in the language of the input prompt.
Two constraints are recorded explicitly in the documentation: generation runs as a single turn, and iterative multi-step editing is not supported in the architecture; and the filtering layer blocks requests that target a specific artist voice or copyrighted lyrics. Every output carries a SynthID audio watermark, which Google states is imperceptible to human hearing.
What the eight-cent rate means
This rate means cost scales with the number of songs produced, not with minutes and not with tokens. A hundred songs comes to eight dollars. For comparison: Eleven Music consumes 900 credits per minute, which on the $6 plan works out to roughly 33 minutes a month, while Suno and Flow Music are billed fundamentally by subscription. This is the one figure in this tier that maps directly onto the unit of work.
The boundary separating it from Flow Music
Flow Music is Google consumer product, and its landing page identifies its underlying model as Lyria 3.5. No such model is recorded in Google API documentation, and no leaderboard row carries that name. So the figures on this page belong to this model, not to that product, and this report does not attribute them to that product.
In practice the two are separate deployment paths: this one is invoked through an API and its user is a developer; the other ships an interface, a library, and video, and its user is someone who wants to produce a song.
Reachability from Iran
Iran is not recorded on the list of regions where the Gemini API is reachable, a finding pulled directly from Google official page. There is no free layer in this tier to offset that either: the official pricing page flags both Lyria rows as not available in the free column.
Using it from Iran
This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.
- Reachable
- blocked
- Payment
- no working route
- Free tier
- no
- Measured on
How we checked: Google officially publishes the list of regions where the Gemini API is reachable, and Iran is not recorded on that list; the same source has been used as the measurement reference for the other Google API models in this reference. For the music model tier, no free layer is provisioned in the pricing architecture at all: the official pricing page flags both Lyria rows as not available in the free column.
Documented strengths
- A transparent per-song pricing rate, a unique feature in this tier
- 44.1 kHz stereo output with WAV available on the Pro model
- A SynthID watermark on every output
- Fifth position on the instrumental leaderboard, above Eleven Music and Udio
Room for improvement and technical considerations
- Maximum output length is not specified as a figure; the documentation only records a couple of minutes.
- Generation runs as a single turn and iterative multi-step editing is not supported in the architecture.
- No free layer exists in the pricing configuration, unlike the rest of the Gemini models.
- On the vocals leaderboard it sits thirty five points below its position on the instrumental leaderboard.
- Iran is not recorded on the list of available regions.
Technical verdict
For a software workload that must produce music, this model carries the cleanest pricing structure in this tier and cost computation is trivial. For a user who wants to sit down and build a song, this model is not the right tool, and the consumer products are.
Frequently raised questions, with documented answers
What is the pricing rate for Lyria 3 Pro
Eight cents per full song. The Clip model, which produces thirty seconds, is billed at a four-cent rate. Neither carries a free layer.
Is Lyria the underlying model behind Flow Music
Not exactly. Flow Music identifies its underlying model as Lyria 3.5, and no such model is recorded in Google API documentation. The figures on this page belong solely to lyria-3-pro-preview.