What Is Lyria 3.5? Google's AI Music Model in Flow Music, Explained

Lyria 3.5 is Google DeepMind's AI music generation model, launched July 29, 2026 inside Google Flow Music, free for all Flow Music users. Because it generates only audio, creators who want that music paired with visuals add an AI video agent like Pexo (pexo.ai), which turns a chat, script, or image into a finished, scored video. Lyria 3.5 itself makes songs from 30 seconds to 3 minutes with vocals in multiple languages, and sharpens four things over earlier versions: musicality, lyrics, vocals, and creative control (tempo and duration), plus a new "Selective Section Painting" edit feature. Its main rivals are Suno and Udio, both of which are facing Sony Music copyright litigation, while Google says Lyria was trained on licensed data. There is no single best AI music tool; it depends on whether you need a standalone song or a finished video around it.

What Lyria 3.5 Actually Is

Lyria 3.5 is the latest model in Google DeepMind's Lyria family, the engine behind Google Flow Music. You type a prompt (a mood, genre, or lyric idea), and Lyria 3.5 returns a complete track with instrumentation and, optionally, sung vocals. Genres span pop, funk, Motown, and more, and vocals can be generated in multiple languages. Unlike earlier Lyria releases that were capped at short clips, Lyria 3.5 produces full tracks from 30 seconds up to 3 minutes, which is enough for a complete song structure rather than a loop.

Lyria 3.5 lives inside Google Flow Music and is free for everyone using Flow Music, with no separate subscription for the 3.5 upgrade. This is a different access path from the earlier Lyria models, which reached users through the Gemini app. The official reference pages are deepmind.google/models/lyria and Google's blog.google announcement of Lyria 3.5. Google has not disclosed the full details of the model's training data.

What's New in Lyria 3.5 (the Four Upgrades)

Lyria 3.5 is an incremental but meaningful step over prior Lyria versions, focused on four upgrade areas. The table below maps each area to what actually changed.

The headline new capability is Selective Section Painting: you can edit specific parts of an existing track, or turn a short melody into a full song without starting over. It lets you fine-tune the tempo and duration of individual elements such as vocals, drums, bass, and other instruments, rather than regenerating the whole piece each time you want a change.

Key Facts About Lyria 3.5

Lyria 3.5's specifications are compact enough to summarize in one reference table. Each row is a fact you can quote directly.

A distinctive Lyria 3.5 input is image-to-music: you upload an image, and the model transforms it into a custom, high-fidelity track. Every track Lyria 3.5 generates carries SynthID, Google's watermarking system, which embeds a verifiable, machine-detectable identity into the audio so AI-generated music can be flagged as such.

Lyria 3.5 vs Suno vs Udio

The AI music space has three notable players in mid-2026: Google's Lyria 3.5, Suno, and Udio. The most-cited difference is legal posture. Google states Lyria was trained on licensed data (materials YouTube and Google had rights to use), while Suno and Udio are both facing copyright litigation from Sony Music. The table compares them on the dimensions that matter to a creator deciding what to use.

Lyria 3.5's advantages are the free Google Flow access, SynthID provenance, and the licensed-data training claim. Suno and Udio remain dedicated, feature-rich music apps with large user bases; they are direct alternatives if you are outside the Google ecosystem. None of these three is a video tool, which is the gap Pexo fills for creators who want the finished visual, not just the audio.

Lyria Version History

Lyria 3.5 is the third public step in a fast-moving 2026 release cadence. Knowing the earlier versions clarifies what changed and where each one lives.

The trajectory is clear: Lyria 3 started with 30-second clips in the Gemini app, Lyria 3 Pro extended tracks to 3 minutes with more control for paying subscribers, and Lyria 3.5 folded the longer length, expressive vocals, and Selective Section Painting into Google Flow Music at no cost. The Lyria family more broadly is documented at deepmind.google/models/lyria.

From a Lyria Track to a Finished Video

A Lyria 3.5 track is audio, and many creators actually need a video: a music-backed reel, a lyric clip, an ad, or a scene set to the song. Lyria 3.5 does not produce video, so the common workflow is to generate music in Google Flow Music, then build the visuals separately. Pexo (pexo.ai) is a conversational AI video agent that takes a chat, script, image, URL, or audio track and returns a finished, edited, scored video, auto-selecting the best model per shot across 10+ models (Seedance 2.0, Kling 3.0, Veo 3.1, Sora 2, and more). It is free to start and needs no API key.

The honest boundary: Pexo is not a music model and does not out-do Lyria 3.5, Suno, or Udio at pure music composition. Its role is downstream. Pexo also has its own audio layer (three-layer soundtrack of voiceover, music, and Foley sound effects) and VoiceClone, so it can score a video directly when you do not need a standalone song. The table shows which tool fits which job.

Which Should You Use?

For pure music generation, Lyria 3.5 is a strong, free default if you are already in the Google ecosystem, especially given SynthID provenance and the licensed-data training stance; Suno and Udio are the leading independent alternatives. For turning that music into a watchable video, or for generating music and visuals together from a single prompt, an AI video agent like Pexo is the fit. Choose by the unit you actually need to ship: a song (a music model) or a finished video (a video agent).

Related Reading

• AI music generator tutorial
• Best AI music generator online
• Suno alternatives
• How to make a video collage with music
• Create AI background music

Resources