Google Launches Lyria 3.5
Google expands its Lyria 3.5 music generation model to the Gemini app and API.
Anonymous
Key Points
1
Lyria 3.5 is available globally to all Gemini users on the web and in the mobile app, not only through developer and creator tools.
2
The Gemini API generates 44.1 kHz stereo audio and allows developers to submit up to 10 images, write lyrics using section tags, and use timestamps to specify when instruments should enter.
3
Lyria 3.5 generation is single-turn: users cannot iteratively edit an already generated clip through follow-up prompts, and Google lists the older Lyria 3 Clip separately as a model that always produces 30-second tracks.
4
The Munich court held Suno responsible for infringing outputs rather than accepting the argument that complex prompts transferred responsibility to users. It also asserted jurisdiction over training conducted outside the European Union and rejected the company’s reliance on U.S. fair-use principles.
5
Google DeepMind says Lyria 3.5 can generate a requested duration from a 60-second clip to a full three-minute song, lets users write lyrics or generate them around themes, and can turn an uploaded image into a track.
Google has rolled out Lyria 3.5, its advanced AI music-generation model, to all Gemini users worldwide on web and mobile apps, plus to developers through the Gemini API. The Decoder reports the new model generates full songs up to three minutes long with more expressive vocals, richer arrangements, and improved lyrics compared to its predecessor. Users can craft tracks from text prompts, images, or custom lyrics by selecting genres and vocal styles.
The expansion marks Google's answer to competitors like Suno, which WebProNews notes faces mounting legal pressure after a Munich court ruled it liable for copyright-infringing outputs. Google has built in SynthID watermarking and safeguards against generating voices that mimic specific artists or reproduce copyrighted lyrics, signaling the company's attempt to navigate the music industry's growing concerns about AI-generated content.
How Lyria 3.5 Works for Regular Users
Gemini users can now generate full-length songs directly within the app without specialized coding knowledge. Jo24 explains that the model turns simple text prompts into complete tracks with verses, choruses, and bridges. Users pick a genre, choose between vocal or instrumental styles, and let Lyria compose. The entire process happens in a single generation — users cannot edit a finished track through follow-up prompts like they might with image tools.
Tracks generated on Lyria 3.5 stretch up to three minutes in length, a major jump from the 30-second clips produced by the older Lyria 3 model. The Decoder reports users can also upload images to transform them into musical compositions, adding a creative visual-to-audio pipeline to Gemini's toolkit.
Developer Tools Grant Granular Creative Control
The Gemini API gives developers and creators far more precision than casual users get. Trividi Digital notes developers can write custom lyrics using section tags, upload up to 10 images to guide musical direction, and use timestamps to specify when individual instruments should enter or exit. The API outputs high-fidelity stereo audio at 44.1 kHz, the standard for professional music production.
Creators working in Google AI Studio, Flow Music, or Google Vids can control lyrics, tempo, duration, and instrument timing all within their existing workflows. This integration lets filmmakers and content producers layer AI-generated soundtracks directly into their projects without switching between separate tools.
Copyright Safeguards Amid Industry Lawsuits
Google built two core protections into Lyria 3.5 to reduce copyright risk. SynthID watermarking embeds invisible fingerprints into generated audio, letting rights holders identify AI-created tracks. The model also refuses requests to mimic specific artists' voices or reproduce copyrighted lyrics, blocking users from directly recreating protected work.
WebProNews reports a Munich court recently held Suno liable for infringing outputs even when users supplied the prompts, rejecting Suno's argument that complex user instructions transferred legal responsibility to them. The court also asserted jurisdiction over training data gathered outside the EU and dismissed Suno's reliance on U.S. fair-use principles. Google has not stated the ruling directly shaped Lyria's restrictions, but the legal landscape clearly influenced the company's cautious approach.
Sentiment
Mostly PositiveNEGATIVENEUTRALPOSITIVE
Publishers
12
Articles
18
Reach
30
A
AnonymousAuthor
