Lyria¶
★★★★★ Basic
Development line: project:lyria · thread lyria-development
Last event: - · 0 dated since - · Researched: 2026-09-05 · confidence: high
What it is¶
Lyria is Google DeepMind’s music-generation system for creators and developers. Lyria 3.5 is available through Flow Music, Gemini, Gemini API, AI Studio, and Google Vids.
- Text prompts produce music and lyrics.
- Generated tracks carry SynthID watermarking.
Current product messaging supports tracks from short clips to up to three minutes. Use it for original, prompt-directed music, then review every output before release.
Development line¶
- The dated line is not written up yet; what is known stands in the sections below.
What changed¶
- 2026-02-18 — Lyria 3 entered the Gemini app beta for 30-second text- or image-directed tracks.
- 2026-03-25 — Google opened Lyria 3 and Lyria 3 Pro developer preview in Gemini API and AI Studio. Clip targeted 30-second assets and Pro targeted songs up to about three minutes.
- 2026-07-29 — Lyria 3.5 launched in Google Flow Music with claimed improvements to melodic structure, lyric adherence, vocal expression/pronunciation, and tempo/duration control.
- 2026-07-30 — Flow Music’s linked launch corresponds to Lyria 3.5, not a separate Lyria model or a first launch. Google’s primary announcement is dated 2026-07-29, while Google Korea published its local announcement on 2026-07-30.
- 2026-09-04 — Lyria 3.5 became available globally in Gemini web and mobile, plus Gemini API, AI Studio, and Google Vids.
How to use this¶
As of 2026-07-30, do not treat the Flow Music URL alone as a documented Lyria development event. Research the linked resource and the underlying post before relying on it.
- Open Google Flow Music or Gemini, sign in, and confirm that the music-generation surface is available for the account. — https://www.flowmusic.app/
- Choose a genre and vocal or instrumental direction, then specify the intended use, tempo, duration, instrumentation, and lyrical direction in the prompt. — https://blog.google/innovation-and-ai/products/gemini-app/better-tracks-lyria-gemini/
- For an integration, use Gemini API or AI Studio. Use text mode for a track brief or composer mode to direct sections such as intro, verse, and bridge. — https://blog.google/innovation-and-ai/technology/developers-tools/lyria-3-developers/
- Listen through the generated track and check that it fits the intended project before using it. — https://deepmind.google/models/lyria/
Best practices¶
- Describe concrete musical controls—tempo, key, dynamics, vocal style, and sectional progression—rather than relying only on a genre label. — https://blog.google/innovation-and-ai/technology/developers-tools/lyria-3-developers/
- Use the shorter clip-oriented option for rapid prototypes and social assets. Reserve full-song generation for longer structured work. — https://blog.google/innovation-and-ai/technology/developers-tools/lyria-3-developers/
- Treat generated music as a draft: review whether the track matches the creative brief before publication. — https://deepmind.google/models/lyria/
Superseded by this¶
- 2026-02-18 — Lyria 3’s Gemini-app 30-second-track limit is superseded for current consumer use by Lyria 3.5’s short-or-longer-track options.
- 2026-07-29 — Lyria 3 guidance is superseded where Lyria 3.5 is available. Use the newer model for improved lyrics, vocals, musicality, and tempo/duration control.
Still unknown¶
- The required response schema has no event_findings or new_events fields, so dated additions and correction are preserved in what_changed rather than emitted as separate fields.
- The Flow Music homepage yielded no readable product text in the retrieved page; it is included only as the supplied product URL and current access surface.
- Google’s public materials do not provide independent benchmark scores or a complete public training-data inventory for Lyria 3.5.
Sources¶
| source | title | read |
|---|---|---|
| https://www.flowmusic.app/ | Google Flow Music | 2026-09-05 |
| https://blog.google/innovation-and-ai/models-and-research/google-labs/lyria-3-5/ | Introducing Lyria 3.5 in Google Flow Music | 2026-09-05 |
| https://blog.google/intl/ko-kr/company-news/technology/lyria-3-5-kr/ | Lyria 3.5 launch in Google Flow Music | 2026-09-05 |
| https://deepmind.google/models/model-cards/lyria-3-5/ | Lyria 3.5 Model Card | 2026-09-05 |
| https://blog.google/innovation-and-ai/products/gemini-app/better-tracks-lyria-gemini/ | Create your best tracks yet with Lyria 3.5 in Gemini | 2026-09-05 |
| https://blog.google/innovation-and-ai/products/gemini-app/lyria-3/ | A new way to express yourself: Gemini can now create music | 2026-09-05 |
| https://blog.google/innovation-and-ai/technology/developers-tools/lyria-3-developers/ | Build with Lyria 3, our newest music generation model | 2026-09-05 |
| https://deepmind.google/models/lyria/ | Lyria 3.5 | 2026-09-05 |
Agent brief¶
- Subject:
project:lyria, threadlyria-development, 0 dated events - → -. - Practical note: As of 2026-07-30, do not treat the Flow Music URL alone as a documented Lyria development event. Research the linked resource and the underlying post before relying on it.
- Confidence: high. Dated supersedes above are the authority for what is obsolete.