Lyria 3.5 in Gemini Puts Full-Length Song Generation in Every User's Hands

Google has opened Lyria 3.5 in Gemini to every user of its assistant, turning a music generator that spent the summer inside a specialist tool into a button in the prompt bar used by hundreds of millions of people. The model went live in the Gemini app and the Gemini API on September 4, free on the free tier on web and mobile in every country where Gemini is available. Tracks now run two to three minutes, replacing the 30-second clips free users had been limited to since February.

Google calls Lyria 3.5 its best-sounding music model. It generates vocals, timed lyrics and complete instrumental arrangements from a written description or an image prompt, and it lets users select a genre or describe one, switch between vocal and instrumental versions, and start from templates for short or long pieces. What comes back is a structured song with verses, choruses and bridges rather than a looped clip.

Reach is where the change sits. Outside the Gemini app, the same model runs in Google Flow Music, Google AI Studio and Google Vids, and developers can call it through the Gemini API under the identifier lyria-3.5, classified as a preview release.

What Lyria 3.5 in Gemini Actually Changes

The model itself is not new. Google shipped it first inside Flow Music on July 29, then extended it to the assistant and the developer interface five weeks later. The September 4 move is a distribution event, and the difference shows in the numbers.

Two entries in that list matter more than they look. The API returns 44.1 kHz stereo files in MP3 and WAV, the same format and sample rate streaming services and video editors expect, so generated audio drops into an existing workflow without conversion. The timed-lyrics feature aligns vocals to a timeline, which is what makes a generated track usable underneath a video rather than merely listenable on its own.

For developers, the preview label cuts both ways. It signals that Google will keep changing the endpoint, so applications built on lyria-3.5 carry integration risk until the model graduates to general availability. Against that, a per-song rate near eight cents makes an audio feature affordable for apps that could never justify licensing a music catalogue.

Read together, the free consumer tier, a sub-ten-cent developer rate and a three-minute ceiling describe a product built for volume. A video editor who needs a backing track, a small business after a jingle, or anyone who wants a custom ringtone can generate one inside an app already open on their phone.

The Trade-Offs: Free Reach Against Rights and Detection

Making the model free at scale solves a distribution problem and creates a provenance one. Every Lyria 3.5 track carries Google's SynthID watermark, an embedded marker designed to survive compression and be read by detectors. Volume makes that marker matter: it is the main way a streaming platform, a rights body or a listener can tell generated audio from recorded audio.

Watermarking answers a narrower question than it appears to. SynthID records that a track came out of Google's model. It does not establish who owns the composition, whether the output can be monetised, or what happens when a generated track closely resembles an existing recording. Those questions sit with contracts and courts.

Timing sharpens the point. The rollout lands while disputes over how music models are trained and how services should treat synthetic uploads are still unresolved, including litigation involving the collecting society GEMA and the generator Suno, and moves by streaming services such as Deezer to flag AI-generated material. A free tier reaching every Gemini user pushes far more generated audio into that unsettled system than Flow Music ever could.

Competitive pressure runs in both directions. Specialist generators such as Suno and Udio have built businesses on subscriptions for song creation. Bundling a capable model free into an assistant people already open for search, writing and image work removes the main reason a casual user would pay for a separate music tool, even if dedicated platforms keep advantages in editing, control and catalogue features.

What Creators Should Weigh

For individuals, the arithmetic is simple. Generation costs nothing inside Gemini and eight cents per song through the API, so the barrier is the prompt rather than the budget. Short-form video creators, podcasters and hobbyist musicians can test ideas at a scale that paid credits discouraged.

Musicians face a different calculation. A model that writes a finished arrangement from a sentence competes with paid studio work, and a free tier removes the price signal that specialist tools used to provide. The professional case for hiring a human rests on rights clarity, editability and the ability to license the result, none of which the rollout addresses.

Three limits deserve attention before anyone plans around the model. The API carries a preview classification, so interfaces and terms can shift. The three-minute ceiling rules out anything close to a release-ready arrangement. The announcement covers capability and watermarking rather than ownership, leaving the commercial rights attached to a generated track an open question for creators to settle before publishing.

Google has not detailed how it will handle the volume a global free rollout produces. What is clear is that the entry point for AI music has moved from a specialist subscription to a chat window, and that provenance is now the contested ground.

Why This Matters

The Lyria 3.5 in Gemini launch shifts the practical question for creators from whether AI can write a song to who owns the result and how platforms identify it. For Google, giving the model away inside Gemini repeats the pattern it used for image generation: make a frontier capability ordinary and let rivals sell the specialist version. Users who wanted a soundtrack without a budget gain the most, while subscription music tools and the rights infrastructure that has to sort recordings from generated audio absorb the pressure.

Sources

Create your best tracks yet with Lyria 3.5 in Gemini.

Related Articles

• Lyria 3.5 music generation opens to everyone in Gemini app and API
• Google Expands Gemini Ecosystem with Native macOS App and Lyria Music Tools
• Google Launches Gemini 3.1 Flash TTS for AI Audio

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.