Lyria 3 Arrives in Gemini: Google Turns Text Prompts Into 30-Second Music Tracks

Google Puts a Recording Studio Inside a Chatbot

On February 18, 2026, Google rolled out Lyria 3 inside the Gemini app, making it the first major AI assistant to offer built-in music generation. Users can now type a text prompt, attach an image, or even share a video reference, and receive a 30-second music track complete with instrumentals, vocals, and auto-generated lyrics. The feature is available to all users aged 18 and older in eight languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.

Lyria 3 is the third generation of Google DeepMind's music synthesis model. Unlike earlier versions, which required users to supply their own lyrics, Lyria 3 automatically generates lyrics based on the prompt. The model also offers control over musical style, vocal characteristics, and tempo, producing what Google describes as "more realistic and musically complex tracks" than its predecessors.

How Lyria 3 Works in Practice

The workflow is straightforward. Inside the Gemini app, users select the music generation option and enter a prompt describing the desired track. A prompt like "upbeat electronic track with female vocals about a summer road trip" produces a 30-second clip within seconds. Users can also attach an image, such as a sunset photograph, and ask Gemini to create a soundtrack that matches the mood.

Google has built in iterative refinement. After hearing an initial generation, users can ask Gemini to adjust the tempo, change the vocal style, or shift the genre. This conversational approach differentiates Lyria 3 from standalone music AI tools like Suno or Udio, which operate outside a general-purpose assistant context.

The model generates complete compositions rather than isolated loops or stems. Each track includes a structured arrangement with intro, verse, and chorus elements, reflecting Google's investment in understanding musical form. However, the 30-second limit constrains the output to what is essentially a demo clip rather than a full song.

Copyright Protections and SynthID Watermarking

Google has implemented several safety measures. All tracks generated through Lyria 3 carry SynthID, Google DeepMind's imperceptible watermark for identifying AI-generated content. SynthID embeds an inaudible signal directly into the audio waveform, allowing automated detection systems to identify the content as machine-generated even after editing, compression, or format conversion.

The model is designed for original expression, not mimicry. If a user specifies an artist name in their prompt, Gemini treats it as broad creative inspiration and generates a track that shares a similar style or mood rather than attempting to replicate the artist's voice or signature sound. Google has also built content filters that check outputs against existing copyrighted material.

These measures address the legal challenges that have engulfed competitors. Suno and Udio both face active lawsuits from major record labels alleging copyright infringement. Google's preemptive approach, combining watermarking, style-not-copy generation, and output filtering, positions Lyria 3 as a more legally defensible option for creators and businesses.

Availability and Pricing Tiers

Lyria 3 launched first on Gemini's desktop interface, with mobile app availability rolling out over the following days. All users 18 and older can access the feature, but usage limits vary by subscription tier. Google AI Plus, Pro, and Ultra subscribers receive higher generation limits, while free-tier users face more restrictive daily caps.

Google has not published specific rate limits for each tier. This tiered approach mirrors how Google handles other compute-intensive Gemini features like extended thinking and Deep Research, reserving higher capacity for paying subscribers while maintaining free access as a discovery mechanism.

The feature supports eight languages at launch, with Google indicating plans to expand language coverage and improve audio quality in future updates. The current 30-second track length is also expected to increase as the model matures.

Competitive Context: The AI Music Race

Lyria 3 enters a crowded and contentious market. Suno, which raised $125 million in 2025, generates full-length songs and has built a large consumer following. Udio offers similar capabilities with a focus on audio quality. Meta's MusicGen provides an open-source alternative for developers.

Google's advantage is distribution. By embedding Lyria 3 directly inside Gemini, which serves hundreds of millions of users, Google bypasses the need for users to discover and adopt a separate music tool. A user working on a presentation in Gemini can generate a background track without leaving the conversation. A content creator can prototype a soundtrack alongside other creative tasks.

The integration also signals Google's broader strategy of positioning Gemini as a universal creative assistant. With image generation via Imagen, video understanding, and now music creation, Gemini offers a multimodal creative toolkit that no single competitor currently matches in scope.

Limitations and Considerations

The 30-second track length is the most significant constraint. While sufficient for social media clips, podcast intros, or presentation backgrounds, it falls short for musicians or producers seeking full-length compositions. Google has not committed to a specific timeline for extending track duration.

Audio quality, while improved over Lyria 2, still carries artifacts that distinguish AI-generated music from professional studio recordings. Vocals can occasionally sound slightly synthetic, and complex arrangements sometimes lose clarity in the mix. These are limitations shared by all current music generation models, not specific to Lyria 3.

The style-not-copy approach, while legally prudent, means that users cannot generate tracks that closely emulate specific artists. For creators who want a track "in the style of" a particular musician, the results may feel generic. This is a deliberate design choice that prioritizes legal safety over creative specificity.

Conclusion

Lyria 3 represents a meaningful step in making AI music generation accessible to mainstream users. By embedding the capability directly inside Gemini, Google eliminates the friction of adopting a separate tool and positions music creation as just another capability of a general-purpose AI assistant. The combination of SynthID watermarking, anti-mimicry filters, and copyright checks makes it the most legally cautious music AI on the market. For content creators, marketers, and casual users who need quick, original background music, Lyria 3 delivers practical value today. Professional musicians and producers will find the 30-second limit and audio quality constraints restrictive, but the trajectory is clear: AI music generation is moving from novelty to utility.