Google is expanding its generative music stack again, this time with Lyria 3.5 showing up across Gemini-adjacent surfaces. The most relevant starting point for creators is Google’s official Gemini music-generation page, which explains how Lyria fits into Gemini’s creation flow: https://gemini.google/us/overview/music-generation/.
The headline is not “AI can make music.” The headline is duration plus editability plus packaging. Lyria 3.5 pushes beyond short clips into full songs measured in minutes, with more explicit control over things like tempo and duration. Those are the knobs creators actually need if the goal is “usable in a cut,” not “cool for 15 seconds.”
Big picture: Google is turning Gemini into a media workstation where music is not a separate app, it is a button inside the same place you are drafting scripts, generating images, and building video.
What Google shipped
Lyria has existed as Google and DeepMind’s music model line, but Lyria 3.5 is being framed as a musicality and control upgrade, not a totally new category.
For the clearest public “what is this” grounding on the developer side, Google’s Gemini API documentation lays out the current Lyria lanes developers see, including “clip” and “pro” style generation: https://ai.google.dev/gemini-api/docs/music-generation.
Important nuance: in the Gemini API documentation, Google lists Lyria 3 models for developers, while Lyria 3.5 is marketed and surfaced primarily in Flow Music and Gemini’s consumer-facing experiences.
If you want COEY context on how Google has been integrating Lyria into Gemini as a creator workflow, see our earlier coverage here: https://bluelightningtv.com/2026/02/21/gemini-adds-lyria-3-music-generation-in-app/.
Across documentation and reporting, the practical changes creators will notice cluster into three areas.
Longer tracks, not just snippets
Earlier in-app music features were designed around short cues. In Google’s ecosystem, that is commonly represented as 30-second “clip” experiences, while longer outputs show up via higher-capability surfaces.
With Lyria 3.5 in Flow Music and in higher-capability Gemini surfaces, the model is discussed in the context of songs that can run up to around three minutes. That is an important shift for anyone building:
- YouTube intros and outros that are not loops
- Branded tracks that need verse and chorus structure
- Longer explainers where music needs to evolve across sections
Better control over time
Two controls matter more than most people admit: tempo (BPM) and duration. Lyria 3.5 is being positioned as an upgrade here, aligning with Flow Music coverage emphasizing tempo control and improved musicality: https://www.androidcentral.com/apps-software/set-the-tempo-googles-lyria-3-5-debuts-with-updates-to-ai-musicality-in-flow-music.
Creator packaging: audio exports and shareable video wrapper exports
One detail that matters in real workflows: on some Gemini and Flow Music surfaces, the tool can export not only a downloadable audio file, but also a shareable video wrapper (typically the audio paired with a static or simple visual).
For more detail on how Gemini’s music creation can be packaged and shared in the consumer experience, Google’s Gemini Help article covers the workflow: https://support.google.com/gemini/answer/16901237.
What Lyria 3.5 does best
This is news, not a pep rally, so here is where the value actually lands.
Fast originals for content packaging
Lyria 3.5 is most useful when music is a production component, not the product. Think:
- Ad variants that need different energy levels
- Mood beds under voiceover
- Fast identity music for a recurring series
- Quick drafts for client review where “directionally right” beats “perfect”
The win is speed without stock-library deja vu.
Structured songs with vocals (sometimes)
In Flow Music, Lyria 3.5 is reported and marketed as supporting lyrics and vocals, with improvements in vocal expressiveness and lyric handling. Availability can vary by surface and settings, and instrumental generation remains a common default workflow.
Reality check: Lyria 3.5 is closing the gap between “demo” and “usable,” but it is not replacing a songwriter’s taste. It is replacing the blank page and the licensing scavenger hunt.
Multimodal vibe steering
Google has been leaning into multimodal prompting across Gemini, and Lyria benefits from that direction. Depending on the surface, prompts can be guided beyond plain text, for example with images for vibe and style. The practical value is reducing misfires when you are trying to match a visual brand.
What it does not do (yet)
A lot of early chatter around Lyria 3.5 lumps in “full music video generation.” Based on what is publicly documented and consistently demonstrated, the safer framing is:
- Lyria generates music
- Some Gemini and Flow Music surfaces can package and export it in creator-friendly ways, including video wrappers
- Gemini’s broader ecosystem includes video generation tools, but Lyria itself should not be treated as a guaranteed prompt-to-cinematic-music-video engine
So if you are imagining “one prompt to three-minute song to perfectly synced narrative video,” treat that as an aspirational workflow, not a guaranteed built-in feature today.
Lyria model landscape
To keep it clean, here is how the current Lyria naming tends to map to creator outcomes in Google’s ecosystem and docs.
| What you need | What to use | What you get |
|---|---|---|
| Quick idea loops | Lyria clip-style experiences | Short tracks commonly around 30 seconds for drafts and tests |
| Full-ish song drafts | Lyria pro-style experiences and Flow Music (Lyria 3.5) | Longer structured songs up to about three minutes depending on surface |
| Share-ready post file | Export options in the product surface | Audio download and, on some surfaces, a shareable video wrapper |
This is the shift: Google is productizing music generation as an output type, not a novelty demo. That is why duration, tempo, and export formats suddenly matter more than “wow, it made a beat.”
Why this matters for creators now
Lyria 3.5 is part of a bigger platform move: Gemini is becoming a creator stack, not a chatbot. Music sits next to image generation, writing, and in other parts of the ecosystem, video generation.
That changes workflows in three practical ways.
More iterations, earlier
Instead of picking music at the end, you can test music while the cut is still a draft. That means pacing decisions, especially for short-form, get better faster.
Less one-person bottleneck
One creator can now rough in concept to script to thumbnail copy to music bed to exportable asset, without recruiting a separate music person for early drafts.
Sonic branding gets cheaper to explore
Not cheap to perfect, but cheap to explore. Teams can iterate toward a signature vibe without paying the “every test costs money” tax.
What to watch next
If Google keeps pushing Lyria forward, the next meaningful upgrades will not be marketing adjectives. They will be operational:
- More predictable structure control, like intro, verse, chorus timing you can trust
- Better editing primitives, like section-level regeneration
- Tighter pairing with video workflows, beyond export wrappers toward real sync points
For now, Lyria 3.5’s news value is simple: longer, more controllable AI music inside the same assistant creators already use, with outputs that can be shared quickly. That is not hype. That is a workflow advantage.
