Song ideation
Writers and producers exploring melodies, arrangements, lyric structures, vocal character and genre directions.
Independent tool overview
Lyria 3.5 at a glance
Lyria 3.5 is Google DeepMind's current text-to-music model, launched in Google Flow Music with improved musicality, lyric structure, vocal realism and pronunciation, plus better tempo and duration control. It can create coherent songs with instruments, lyrics and vocals and is part of a broader Lyria family that supports image-conditioned music and tracks up to roughly three minutes. Google Flow Music has a free plan and paid plans from $8 to $64 per month, while the Gemini API separately exposes Lyria 3 Clip and Pro preview models at $0.04 and $0.08 per request. Lyria is useful for ideation, demos and licensed production workflows, but every song still needs rights, originality, lyric, voice, mix, metadata, distribution and human review.
Visit the official Lyria 3.5 site ↗.jpeg&w=3840&q=75)
- Developer
- Google DeepMind
- Named release surface
- Google Flow Music
- Primary input
- Text prompt; the wider Lyria family also supports image-to-music workflows
- Outputs
- Music audio plus generated or user-directed lyrics and vocals
- Headline improvements
- Musicality, lyric structure, vocal realism, pronunciation, tempo and variable duration
- Track length
- Variable; Google describes cohesive Lyria tracks up to about three minutes
- Watermark
- All generated tracks carry an imperceptible SynthID audio watermark
- Flow Music eligibility
- Age-verified users 18 or older in supported regions
- Flow Music price
- Free core plan; paid plans shown at $8, $24 and $64 per month
- Flow Music exports
- Platform comparison lists MP3, WAV, M4A and stem downloads among its features
- Developer API distinction
- Current Gemini API IDs are Lyria 3 Clip Preview and Pro Preview, not a 3.5-specific ID
- API price
- $0.04 per 30-second Clip request and $0.08 per Pro full-song request
- Data review
- Flow Music says a sample of uploads and outputs may be read, annotated or processed by human reviewers
- Reviewed
- August 31, 2026 from current Google DeepMind, Flow Music, Gemini API, Google One and privacy sources
Overview
What Lyria 3.5 is
Google introduced Lyria 3.5 in July 2026 as its most capable music-generation model to date. The model card describes latent diffusion over temporal audio latents with text input and music plus lyric output. Its launch surface is Google Flow Music.
Compared with Lyria 2, Google's evaluations report improved audio fidelity and better adherence to simple and complex lyric instructions. The release also emphasizes richer melodic structure, more expressive and realistically pronounced vocals, and easier control over tempo and song length.
The wider Lyria family supports instrumental or vocal tracks across genres and languages, user-supplied or generated lyrics, image-to-music composition, and high-fidelity exports. Google's family page says cohesive tracks can run up to three minutes, but exact output length and availability depend on the selected surface and model.
Google Flow Music is a separate creative environment for songs, playlists, music videos, projects and collaborative-style production tools. It requires an age-verified user of at least 18 in a supported region. Core features are available at no charge, while the latest models and higher generation limits can require a paid Flow Music or Google AI plan.
The current Gemini Developer API documentation exposes Lyria 3 Clip Preview for fixed 30-second MP3 clips and Lyria 3 Pro Preview for songs lasting a couple of minutes. Google does not label those model IDs as Lyria 3.5, so this page treats the API as a related Lyria 3 developer path rather than claiming it is the named Flow Music release.
All Lyria-generated audio is described as carrying an imperceptible SynthID watermark. Google also applies prompt and output safety controls, including blocking requests for specific artist voices or copyrighted lyrics in the current API documentation. Watermarking and filters do not perform music clearance.
Google Flow Music's privacy notice says uploads, generated outputs, feature usage and feedback may help improve products and machine-learning technologies, and a sample may be reviewed by people. Do not upload confidential, unreleased or rights-restricted material without an approved account, contract and data-handling decision.
Use cases
Who Lyria 3.5 is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Demo creation
Turning a brief or draft lyric into a listenable reference before human performance and production.
Background music
Generating project-specific cues after confirming plan rights, originality and the distribution platform's rules.
Sonic mood boards
Testing tempo, instrumentation, dynamics, texture and emotional direction for a campaign, film, game or product.
Image-inspired composition
Transforming an authorized photo, artwork or visual concept into a musical starting point.
Developer prototypes
Apps that need inexpensive short or full-song generation through the distinct Lyria 3 preview API.
Capabilities
Core Lyria 3.5 features
Full-song generation
Creates coherent instrumental or vocal music with sections, arrangement and an ending rather than only isolated loops.
Improved vocals
Lyria 3.5 targets more realistic, emotionally nuanced singing with better pronunciation.
Lyric generation
Writes lyrics around a theme with improved prompt and structural adherence.
User-supplied lyrics
Allows creators to provide lyrics and describe lead and backing-vocal behavior.
Musical controls
Accepts direction for genre, instruments, tempo, dynamics, density, mood, vocal character and duration.
Global genres and languages
Google demonstrates music across varied genres and vocal languages, subject to pronunciation and cultural review.
Image-to-music
The wider Lyria experience can interpret an uploaded image as a source for musical mood and composition.
Flow Music production space
Places songs inside projects, sessions, playlists, publishing and music-video workflows.
Audio and stem downloads
The Flow Music plan comparison lists MP3, WAV, M4A and stem downloads as platform features.
SynthID watermarking
Embeds an imperceptible signal intended to help identify Google AI-generated or edited music.
Safety filtering
Blocks prohibited prompts and, in the current API, requests for specific artist voices or copyrighted lyrics.
Related API models
Offers Lyria 3 Clip and Pro preview models through Google's Interactions API for programmatic MP3 generation.
Process
How the Lyria 3.5 workflow works
Step 1
Define the rights context
Decide whether the track is a private experiment, client demo, commercial release, advertisement, game, film, stream or public performance and identify the rules that follow.
Step 2
Use original source material
Provide lyrics, melodies, recordings, images and references you own or are licensed to use; avoid copyrighted lyrics, samples and requests to clone a real artist.
Step 3
Write a musical brief
Specify purpose, audience, genre, mood, tempo, meter, key range, instruments, song structure, duration, language, vocal attributes, dynamics and exclusions.
Step 4
Prompt attributes, not imitation
Describe musical traits such as era, instrumentation, production texture and vocal register instead of naming a living performer or demanding an exact copyrighted song.
Step 5
Prototype short directions
Compare several distinct concepts before spending credits or time on full songs and exports.
Step 6
Lock the lyric first
Verify meaning, rhyme, meter, pronunciation, translation, factual claims, cultural references, trademark use and prohibited or defamatory content.
Step 7
Review musical similarity
Compare hooks, toplines, chord progressions, lyrics, arrangement, timbre and vocal identity against known work and run appropriate catalog or fingerprint checks.
Step 8
Inspect the audio
Listen for artifacts, clipping, phase issues, noise, unnatural breaths, timing drift, mispronunciation, broken transitions, stereo problems and unsafe frequencies.
Step 9
Reproduce with human musicians
For an important release, consider rerecording, rearranging and mixing with accountable contributors so the final work has deliberate human authorship and editable stems.
Step 10
Clear every contributor and asset
Document writers, performers, producers, samples, fonts, artwork, voices, images, model plan, prompts and applicable licenses.
Step 11
Master and make accessible
Apply conventional edit, mix, mastering, loudness, metadata, lyric, caption and seizure or sensitivity checks for the target platforms.
Step 12
Preserve provenance
Keep SynthID intact, label AI involvement where required, and retain prompts, dates, versions, source rights, human changes and approvals.
Step 13
Verify distribution rules
Check the selected Flow Music or Google AI plan, commercial rights, distributor, performing-rights organization, social platform, union, client and jurisdiction before release.
Step 14
Protect unpublished work
Disable improvement settings where appropriate and avoid uploading confidential songs or client references that a reviewer or model-improvement process should not see.
Cost
Lyria 3.5 pricing and free plan
Google Flow Music offers core features at no charge and standalone monthly plans shown at $8, $24 and $64. Its pricing page also showed discounted annual-billing equivalents of $6, $18 and $48 per month; checkout is authoritative for cadence, tax and region. Paid Google AI memberships map to the Starter, Plus and Member Flow Music benefits. The separate Gemini API's current Lyria 3 preview models have no free tier and cost $0.04 per 30-second clip or $0.08 per full-song request.
Flow Music Free
$0/month
For core Flow Music features with daily top-up credits and lower concurrency.
- Daily top-up credits
- Two concurrent generations
- All core features
- Latest-model access can require an upgrade
- Free-plan commercial rights are not stated on the public pricing table; verify before monetization
Flow Music Starter
$8/month or $6/month equivalent annually
Entry paid plan, also included as a benefit of eligible Google AI Plus memberships.
- 3,000 monthly credits, marketed as roughly 600 songs
- Daily top-up credits
- Eight concurrent generations
- All core features
- Google AI Plus help expressly lists commercial-use rights
- Extra credits can be purchased while a paid subscription is active
Flow Music Plus
$24/month or $18/month equivalent annually
Higher-volume creator plan, also included as a benefit of eligible Google AI Pro memberships.
- 10,000 monthly credits, marketed as roughly 2,000 songs
- Daily top-up credits
- 12 concurrent generations
- All core features
- Google AI Pro help expressly lists commercial-use rights
- Unused monthly allocation and top-up rules should be checked in-product
Flow Music Member
$64/month or $48/month equivalent annually
Highest standalone plan, also included as a benefit of eligible Google AI Ultra memberships.
- 30,000 monthly credits, marketed as roughly 6,000 songs
- Daily top-up credits
- 16 concurrent generations
- Member badge, events, merchandise and early feature access
- Google AI Ultra help expressly lists commercial-use rights
- Extra credits can be purchased
Gemini API Lyria 3 Preview
$0.04 clip / $0.08 full song
Related developer models with separate preview IDs and paid API billing.
- lyria-3-clip-preview: fixed 30-second MP3 for $0.04 per request
- lyria-3-pro-preview: a couple of minutes for $0.08 per request
- No free API tier
- Paid-tier content is listed as not used to improve Google's products
- Preview models can change and have more restrictive rate limits
- Google does not label these IDs as Lyria 3.5
Pricing checked . Check current pricing at the source ↗
Assessment
Lyria 3.5 strengths and limitations
Where it stands out
- Current Google flagship for text-to-music with a detailed model card
- Improved vocal expression, pronunciation, lyric structure, musicality and audio fidelity
- Can combine instruments, lyrics and vocals into coherent song-length output
- Prompt controls cover musical and production attributes rather than only a genre label
- Supports a wider creative environment for projects, playlists, music videos, downloads and stems
- Free Flow Music plan lowers the barrier to exploration
- Paid plans publish credits, approximate song volume and concurrent-generation limits
- Inexpensive separate Lyria 3 developer API for 30-second and full-song prototypes
- SynthID is embedded in generated music for provenance
- Google documents filtering, red-teaming, ethics review and known limitations
What to consider
- Lyria 3.5's named launch surface is Flow Music; current Gemini API model IDs are branded Lyria 3 Preview, not 3.5
- Flow Music requires an age-verified user of at least 18 in a supported region
- The free plan may not include the latest models and does not publicly state commercial-use rights
- A generated song can contain generic writing, awkward prosody, mispronunciation, repetitive structure, unnatural vocals or audio artifacts
- Prompt adherence and musical quality remain probabilistic and require listening review
- The same prompt can produce different results and API generation is not deterministic
- Current Lyria 3 API generation is single-turn and cannot conversationally edit an existing clip across calls
- The API Clip model is fixed at 30 seconds and the Pro duration is approximate
- Filters can block artist-voice, copyrighted-lyric and other requests, including some legitimate edge cases
- SynthID identifies Google AI involvement but does not prove ownership, originality, consent or clearance
- Commercial-use permission does not guarantee copyright protection or eliminate third-party claims
- Outputs may resemble existing melodies, hooks, arrangements, lyrics, performances or recordings
- A synthetic voice can create publicity, impersonation, performer, union or deceptive-endorsement risks even without naming an artist
- Images used for image-to-music can contain faces, locations, private information, copyrighted art or confidential campaigns
- Google Flow Music collects uploads, generated output, feature usage and feedback for service and machine-learning improvement unless settings change
- A sample of Flow Music data may be reviewed by humans after being disconnected from the Google Account
- Generated outputs are stored until deletion by default; conversations can be retained up to 18 months and system logs for 30 days
- Reviewed or annotated data may remain up to 18 months even after deletion or disabling improvement
- Paid credits, annual discounts, availability, latest-model access and plan benefits can vary by region or account
- Music distribution still requires rights review, metadata, mastering, platform compliance, provenance and accountable human approval
Compare
Lyria 3.5 alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Content Creator
Suno AI
Choose Suno for a mature consumer song-generation workflow with its own creation, editing, community and distribution ecosystem.
Explore Suno AI →Content Creator
Udio
Choose Udio for another prompt-to-song environment with iterative music creation and a dedicated creator community.
Explore Udio →Content Creator
Eleven Music
Choose Eleven Music for ElevenLabs' combined music-generation, remixing, streaming and creator-payout product.
Explore Eleven Music →Content Creator
Stable Audio 3.0
Choose Stable Audio 3.0 when open-weight deployment and a commercially licensed training-data story matter more than Google's hosted ecosystem.
Explore Stable Audio 3.0 →Content Creator
Adobe Firefly Audio
Choose Adobe Firefly Audio for music and sound generation inside an established commercial creative and editing workflow.
Explore Adobe Firefly Audio →Content Creator
Pika Audio
Choose Pika Audio when you need music alongside speech, sound effects and soundtrack models in one audio family.
Explore Pika Audio →Questions
Lyria 3.5 FAQs
What is Lyria 3.5?
Lyria 3.5 is Google DeepMind's current music-generation model. It turns text prompts into music and lyrics and improves on earlier Lyria versions in musicality, audio fidelity, lyric adherence, vocal realism, pronunciation, tempo control and duration control.
Where can I use Lyria 3.5?
Google launched the named Lyria 3.5 model in Google Flow Music. The wider Lyria family also appears across Google products, while the Gemini API currently exposes separately named Lyria 3 Clip and Pro preview IDs.
Is Lyria 3.5 free?
Google Flow Music has a $0 core plan with daily top-up credits and two concurrent generations. Google's help says upgrading may be required for the latest models, so confirm that Lyria 3.5 is available to the current free account before relying on it.
How much does Google Flow Music cost?
The current standalone monthly prices are $8 for Starter, $24 for Plus and $64 for Member. The page also shows annual-billing equivalents of $6, $18 and $48 per month. Checkout, region and tax are authoritative.
How long are Lyria 3.5 songs?
Google describes variable song lengths and says the Lyria family can create cohesive tracks up to about three minutes. Exact length depends on the model, prompt, product surface and current limit.
Can Lyria 3.5 generate vocals and lyrics?
Yes. Users can ask for generated lyrics or provide their own, specify vocal register and character, and request lead or backing-vocal behavior. Check pronunciation, meaning, originality and performance rights.
Can Lyria turn an image into music?
Google demonstrates image-to-music composition in the broader Lyria experience. Upload only images you have permission to use and remove private or confidential details first.
Does Lyria 3.5 have an API?
The model card says Lyria is available to downstream providers through an API, but current public Gemini developer docs expose Lyria 3 Clip Preview and Lyria 3 Pro Preview IDs. They are not explicitly labeled Lyria 3.5, so verify the model version required by your application.
How much does the Lyria API cost?
The current Gemini API prices are $0.04 per 30-second Lyria 3 Clip Preview request and $0.08 per Lyria 3 Pro Preview full-song request. There is no free API tier.
Can I use Lyria music commercially?
Google's help expressly lists commercial-use rights with paid Flow Music benefits attached to Google AI Plus, Pro and Ultra. You must still follow the full terms and clear lyrics, references, voices, samples, trademarks, publicity rights, distributors and local law. Verify free-plan rights separately.
Does Google own Lyria-generated songs?
Google's general terms say it will not claim ownership over original content generated by services that allow it. That does not establish copyrightability, originality, exclusivity or freedom from third-party rights, and plan-specific terms still govern.
Does Lyria copy artists?
Google says current safety controls block requests for specific artist voices and copyrighted lyrics in the API. No filter can guarantee that a melody, hook, vocal timbre or arrangement will not resemble existing work, so consequential releases need similarity and legal review.
What is SynthID in Lyria?
SynthID is Google's imperceptible audio watermark embedded in Lyria-generated tracks to help identify AI-generated or edited music. Preserve it, but do not treat it as a copyright or clearance certificate.
Is Flow Music data private?
Its privacy notice says Google collects uploads, outputs, usage and feedback and may use them to improve products and machine learning. Human reviewers process a sample. Users can disable Help improve Google Flow Music, but reviewed data may remain up to 18 months. Do not submit confidential work.
Bottom line
Our Lyria 3.5 verdict
Lyria 3.5 is a credible step toward musically coherent, vocal-first AI songs inside a surprisingly full production environment. Google publishes more useful safety and model detail than many consumer music generators, and Flow Music's plans are competitively priced. The strongest fit is supervised ideation and production where creators own the source material and can perform conventional music clearance. The biggest caution is that attractive audio can feel finished long before it is original, rights-cleared, mixed, documented or safe to release—and Flow Music's default improvement and human-review path makes confidential material a poor fit.
Visit Lyria 3.5 website ↗Similar Miscellaneous Tools
View all Miscellaneous AI tools →
Antares
Antares - Cisco's small, open security models that scan code locally
Inkling-Small
Inkling-Small - Thinking Machines' compact open model that rivals full-size version
Sonic 3.6 & Ink 2
Sonic-3.5 & Ink-2 - Cartesia's new top-ranked speech and transcription models for voice agents
Kitesurf
Kitesurf - Cloudflare's lightweight, agent-first browser
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.