Google Lyria 3 Launches in Gemini – AI Music Generation With Lyrics, Style Control, and SynthID Watermarking

AI

by Johnnie BehiriFebruary 20^th, 2026

Subscribe to News - Audio

Google DeepMind has released Lyria 3, its most advanced AI music generation model, directly inside the Gemini app. The tool creates 30-second tracks with auto-generated lyrics, custom cover art, and SynthID watermarking from simple text prompts or uploaded images, and is now available in beta for all users aged 18 and over.

The AI music space has been evolving at a remarkable pace, with tools like Suno and Udio pushing the boundaries of what text-to-music generation can do. We covered that development in detail in our AI music generators overview last year. Now Google is entering the consumer-facing arena more aggressively with Lyria 3, bringing its generative music capabilities out of research labs and into the hands of everyday Gemini users, including filmmakers looking for quick soundtrack solutions.

What Lyria 3 brings to the table

Lyria 3 represents a significant step forward from Google’s earlier music generation efforts. Those who remember the rough-sounding outputs of MusicLM from 2023, which Google has since retired, will find Lyria 3 operating on a different level entirely. The model can now generate complete tracks with vocals, lyrics, and layered instrumentation from a text description alone. There is no need to supply your own lyrics; the system writes them based on your prompt.

Google highlights three core improvements over its previous Lyria models. First, automatic lyric generation based on the user’s prompt. Second, greater creative control over style, vocal characteristics, and tempo. Third, more realistic and musically complex output overall. Users can describe a genre, mood, memory, or even an inside joke, and Lyria 3 will produce a polished 30-second clip. The system also accepts photos and videos as creative input, analyzing the visual content to compose a matching soundtrack.

Each generated track comes with custom cover art created by Google’s Nano Banana image model. Tracks can be downloaded or shared via a link, making them easy to distribute.

Filmmaking implications and creative workflows

For filmmakers, the practical question is whether Lyria 3 can serve as a useful tool in actual production workflows. At 30 seconds per clip, it is clearly not designed for scoring a feature film. Google itself frames the tool as a means of personal creative expression rather than professional music production. The company’s blog post states that the goal is to provide a fun way to express yourself, not to create a musical masterpiece.

That said, there are scenarios where quick AI-generated music clips could prove useful in a filmmaking context. Temp tracks for rough cuts, mood references during pre-production, social media content for film marketing, or placeholder audio for pitch decks all come to mind. The ability to upload a still frame or short video clip and receive a tonally matched soundtrack in seconds is a workflow that could save time during early creative stages.

Competitors like Suno and Udio still hold an advantage when it comes to longer-form output and deeper creative controls. Suno generates multi-minute songs with proper verse-chorus-bridge structure, while Udio (whose team, notably, consists of former Google DeepMind researchers) offers features like prompt strength sliders and negative prompting. Lyria 3’s 30-second limit and more casual positioning suggest Google is targeting a different use case for now.

We explored the broader landscape of AI tools for filmmakers in our 2024 recap, and the trajectory is clear: AI-generated music is moving from novelty to genuine production utility, even if we are not quite there yet with every tool.

SynthID watermarking and audio verification

One area where Google is pushing ahead of competitors is transparency and content provenance. Every track generated through Lyria 3 in the Gemini app is embedded with SynthID, Google DeepMind’s imperceptible watermarking technology for identifying AI-generated content. The watermark is woven into the audio at the point of creation and can be detected later without affecting the listening experience.

Google has also expanded Gemini’s verification capabilities to include audio, in addition to images and video. Users can upload an audio file and ask whether it was generated using Google AI. The system will check for SynthID markers and apply its own reasoning before returning a result. For an industry increasingly concerned about the provenance of synthetic media, this is a welcome development. We have previously discussed how Google embeds SynthID into its Veo 3 video outputs as well, signaling a consistent approach across its generative media tools.

Copyright guardrails and responsible development

Google states that Lyria 3 was developed with attention to copyright and partner agreements. The model is designed for original expression rather than mimicking existing artists. If a user’s prompt names a specific artist, Gemini interprets this as broad creative inspiration and generates a track that shares a similar style or mood rather than attempting to replicate a recognizable voice or sound.

The company also says it has filters in place to check outputs against existing content, though it acknowledges this approach may not be foolproof. Users can report content that may violate rights, and all usage is governed by Google’s Terms of Service and generative AI prohibited use policies. This matters in the context of ongoing industry tension around AI training data; reports from early 2024 indicated that Google had previously trained AI music models on copyrighted recordings before approaching rights holders for licensing. The recent licensing agreement between Universal Music Group and YouTube, which includes guardrails around generative AI content, suggests the industry is working toward clearer frameworks.

YouTube Dream Track integration

Lyria 3 is not limited to the Gemini app. The model also powers YouTube’s Dream Track feature, which allows creators to generate custom soundtracks for YouTube Shorts. Previously available only in the United States, Dream Track is now rolling out to creators in additional countries. The integration gives short-form video creators access to AI-generated lyrical verses and instrumental backing tracks tailored to their content.

This positions Lyria 3 as part of a broader Google ecosystem play, connecting the Gemini app for casual music creation with YouTube for content creator workflows. For filmmakers active on YouTube, this could offer a convenient way to generate unique audio for short-form promotional content without licensing concerns.

Availability and access

Lyria 3 is available in the Gemini app for all users aged 18 and over in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. Google plans to expand language coverage over time. The feature is rolling out on desktop first, with mobile app availability following over the coming days. Google AI Plus, Pro, and Ultra subscribers receive higher usage limits, though Google has not specified exact numbers.

For developers and more technical users, Google also offers Lyria through its Vertex AI platform and the Gemini API, though the cloud-based version currently runs on Lyria 2 (lyria-002) rather than the latest Lyria 3 model. That version generates instrumental-only tracks and outputs WAV files.

AI music generation has come a long way since the barely listenable outputs of just two years ago. Lyria 3 is another step forward, though its 30-second limit and casual positioning leave room for dedicated tools in professional workflows. Could you see yourself using AI-generated music for temp tracks, social media, or pitch materials? Let us know in the comments below