Mandopop
Mandarin pop full song
Intended to show natural Mandarin pronunciation, verse–chorus–bridge structure and a rising emotional arc.
Vocals · Mandarin · ~3:00Audio coming soon
LYRIA 3.5 · INDEPENDENT MULTI-MODEL STUDIO
Create full songs with Google Lyria 3.5
Describe the genre, tempo, instruments, vocals and structure you want, then have Lyria 3.5 build a complete track of up to about three minutes — with copy-ready prompts, a tutorial, comparisons and a live point estimate.
- Up to 3 minutes
- Instrumental or vocals
- 8 vocal languages
- Image-to-music
- Live point estimate
Nano Banana Pro is an independent multi-model AI creation platform. It is not the official website of Google, Google DeepMind, Gemini or Flow Music, and has no affiliation with or endorsement from them.
LIVE WORKFLOWGoogle FlowMusic Lyria 3.5
Up to 3 minutesInstrumental or vocals8 vocal languages
01
Lyria 3.5 online music workspace
Available fields, model status and the estimated point cost come from the live catalogue — please confirm them before you submit.
Google FlowMusic Lyria 3.5
AI Music Studio
Estimated cost — points
Model
Loading live model settings…
Generated audio
♫Your generated audio will appear here.
02 · MUSIC DEMOS
Real audio coming soonLyria 3.5 music demo slots
These slots are reserved for finished tracks across languages, genres and use cases. Until real audio is published, every card is explicitly labelled as a placeholder — no fake player is shown.
Cinematic
Cinematic video score
Intended to show how arrangement dynamics track the mood of a scene.
Instrumental · ~1:30Audio coming soon
Commercial
Brand jingle
Intended to show a short hook, a memorable brand tag and a compact song form.
Vocals · Mandarin · ~0:30Audio coming soon
Lo-fi
Lo-fi focus loop
Intended to show a steady groove with soft texture and a loop-friendly ending.
Instrumental · ~2:00Audio coming soon
Electropop
Neon electropop
Intended to show crisp rhythmic placement, layered synths and an expressive vocal.
Vocals · English · ~2:30Audio coming soon
Game OST
Game scene score
Intended to show pacing control and a soundtrack direction usable for concept validation.
Instrumental · ~1:00Audio coming soon
03 · MODEL OVERVIEW
What is Lyria 3.5?
Lyria 3.5 is Google DeepMind’s music generation model. It turns a text prompt into music audio, and Google describes it as the most advanced model in the Lyria family, with variable song lengths up to about three minutes. The launch notes group the improvements into musicality, lyrics and vocal quality: richer melodies and arrangements, lyrics that follow the prompt and the song structure, and more expressive vocal performances.
The same model is reachable from several Google products rather than a single app — Google Flow Music, the Gemini app, Dream Track in YouTube Shorts, Google Vids, Google AI Studio and the Gemini Enterprise Agent Platform. The Lyria family also includes the original Lyria, Lyria RealTime for interactive generation, and the open Magenta RealTime model.
This page is a multi-model creation workspace, not an official Google product. The controls, available fields and point cost you actually see are read from the live catalogue at submit time, so treat the capability notes below as model background and the workspace panel as the source of truth for what you can run right now.
- Input
- Text prompt; official workflows also accept images
- Reference material
- Up to 10 reference images or PDFs per request
- Song length
- Variable, up to about 3 minutes
- Vocal languages
- 8: English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese
- Output
- Music audio, with or without vocals
- Provenance
- Google states tracks carry an imperceptible SynthID watermark
04 · CAPABILITIES
Lyria 3.5 music generation capabilities
Capability notes below are summarised from Google’s own model page and developer documentation. Which fields this site exposes, and what a generation costs, is shown live in the workspace panel.
Full song structure
Organise intro, verse, chorus, bridge and outro for tracks that stay coherent across several minutes.
Richer musicality
Generate more natural melodic movement and deeper arrangement layers, aimed at higher overall fidelity.
Lyrics that follow the brief
Have lyrics written around a theme you describe, or hand over your own lines and have them performed.
More expressive vocals
Steer singer type, range, timbre and delivery, and let the vocal change character as the track progresses.
Detailed creative control
Specify genre and era, tempo in BPM, song key, instruments, mood, intensity and target duration inside the prompt.
Timeline prompting
Lay out timed segments with [mm:ss] markers to script genre shifts or score a video scene by scene.
Image-to-music
Google’s workflows let you upload visual references, including up to 10 images or PDFs, to set the mood of the track.
Negative prompts
Name what to keep out — unwanted vocals, harsh cymbals, distorted guitar — through the negative_prompt parameter.
05 · USE CASES
From full songs to video scoring
Short-form background music
Build a bed that matches the pacing of a story clip, travel edit or social post.
Product and brand audio
Explore ad scores, brand jingles, campaign themes and a sound direction for a launch.
Songwriting sketches
Turn lyric ideas, melodic direction and structure notes into a draft you can play for other people.
Film and game concept scores
Test the musical mood of a character, location, trailer or level before committing budget.
Podcast and show packaging
Produce concept intros, outros, stings and series themes.
Visual to soundtrack
Start from a photo, illustration or moodboard and search for the sound language that fits it.
06 · PROMPT GUIDE
Lyria 3.5 prompt templates
Google’s documentation recommends a structured prompt rather than a pile of keywords. Cover genre and style, mood, instrumentation, tempo and rhythm, vocal style and language, and lyrics; then optionally add arrangement, soundscape and production quality. Every template below can be loaded into the workspace and edited before you submit.
Prompt formula: [Genre & style] + [Mood] + [Instrumentation] + [Tempo & rhythm] + [Vocal style & language] + [Lyrics], then optional [Arrangement/Structure] + [Soundscape/Ambiance] + [Production Quality]
Starting over after the rain
Write a modern Mandarin pop song around 3 minutes long, 92 BPM, in G major. Open with warm piano and acoustic guitar, lead with a clear, emotional female vocal. Structure: intro, verse, pre-chorus, chorus, second verse, bridge, final chorus, outro. The theme is starting over after the rain, and the chorus melody should open up and be easy to remember.
- Tags
- Mandopop, 92 BPM, G major, warm piano, emotional female vocal
- Mode
- Vocals · 180 sec
- Exclude
- heavy distortion, EDM drop, rap
City sunrise score
Write a cinematic instrumental piece around 90 seconds long. Begin with quiet piano and low strings, then bring in French horn, percussion and a wide string section. Follow the mood of a city moving from restrained pre-dawn to a released sunrise. Keep the dynamics natural and the production close to a modern film score. Instrumental, no vocals.
- Tags
- cinematic, orchestral, hopeful, gradual build, instrumental
- Mode
- Instrumental · 90 sec
Neon night drive
Write a retro-futuristic electropop song around 2 minutes 30 seconds, 110 BPM, in B minor. Use an analog synth bass, pulsing arpeggios and a crisp drum machine under a light female vocal. Keep the verses restrained and let the chorus open up with forward drive. Aim for the feeling of driving through a city in the rain at night.
- Tags
- electropop, synthwave, 110 BPM, B minor, female vocal
- Mode
- Vocals · 150 sec
- Exclude
- acoustic folk, growling vocals
Rainy night focus loop
Write a lo-fi hip hop instrumental around 2 minutes, 72 BPM, built on soft jazz piano chords, rounded bass, light vinyl texture and restrained drums. Keep it right for late-night reading and focus work, make the section transitions smooth, and let the ending loop cleanly. Rain in the background, no vocals.
- Tags
- lo-fi hip hop, 72 BPM, jazz piano, rainy night, instrumental
- Mode
- Instrumental · 120 sec
- Exclude
- vocals, hard transients
Summer brand jingle
Write a bright brand jingle around 30 seconds, 118 BPM, with crisp guitar, hand claps, a light bass and short brass accents. Use friendly mixed vocals. The hook should land within the first 5 seconds and leave a clear gap at the end for a spoken brand line.
- Tags
- commercial jingle, 118 BPM, bright, catchy hook, mixed vocals
- Mode
- Vocals · 30 sec
- Exclude
- minor key, melancholy
Deep sea exploration
Write an ambient instrumental piece around 60 seconds for a deep sea exploration scene, in D minor. Layer slowly shifting synth pads, a low pulse, distant underwater textures and sparse piano notes. Keep a sense of calm, the unknown and mild tension, with a wide and spacious image. No vocals.
- Tags
- ambient game soundtrack, D minor, underwater, spacious, instrumental
- Mode
- Instrumental · 60 sec
- Exclude
- drum kit, brass stabs
Gospel-rap build (timeline prompt)
[00:00] Open with a large gospel choir singing an uplifting harmony. [00:15] A heavy modern hip-hop drum beat and a deep bass line drop in under the choir. [00:30] A male lead raps a confident verse while the choir punctuates the lines. [01:00] Move into a triumphant chorus with brass. [01:40] Strip back to a warm organ for an emotional bridge. [02:10] Return at full energy for a final chorus and end on a sustained choir chord at [02:40].
- Tags
- gospel, hip hop, [mm:ss] timeline, choir, brass
- Mode
- Vocals · 170 sec
- Exclude
- lo-fi texture, whispered vocals
07 · TUTORIAL
How to use Lyria 3.5
A five-step path from a loose idea to a draft you can actually listen to.
- 01
Decide the job first
Settle whether you need a full song, a short-form bed, a brand jingle or a game cue, and write down the target length before you prompt.
- 02
Describe the musical skeleton
Work through genre and style, mood, instrumentation, tempo and rhythm, then key and intensity if the section needs them.
- 03
Handle vocals or say you do not want them
Name the singer type, range, timbre, delivery and language — or write "instrumental" to exclude vocals entirely.
- 04
Write the lyrics or ask for them
Put "Lyrics:" before lines you want sung verbatim, describe a theme and let the model write them, or mention where backing vocals should answer the lead.
- 05
Shape the structure and iterate
Use section names or [mm:ss] timestamps to plan the arrangement, check the live point estimate, then change one variable at a time so you can tell what moved.
- 06
Check the result against your intent
Google notes the model is still improving, so listen critically for structure, pronunciation and mix before you use a track anywhere public.
08 · COMPARE
Lyria 3.5 vs Suno: which fits your task?
The table separates publicly documented model capability from what this workspace actually exposes. Account terms, pricing and rights rules belong to each provider and are not interchangeable.
| Item | Google Lyria 3.5 | Suno workflow |
|---|---|---|
| Best fit | Full songs, image-driven scores, product and developer workflows | Songwriting and artist-facing product workflows |
| Text to music | One structured prompt controls genre, mood, instruments, tempo, key and vocals | Prompt-driven; see Suno’s own documentation for its current field list |
| Image or document input | Official workflows accept images, up to 10 images or PDFs | Depends on the Suno product surface in use |
| Song length | Variable, up to about 3 minutes | Depends on the Suno plan and current product limits |
| Vocal languages | 8 languages documented by Google | Depends on the Suno product surface in use |
| Timeline control | Timed [mm:ss] segments for multi-part arrangements | Depends on the Suno product surface in use |
| Provenance | Google states generated tracks carry a SynthID watermark | Follows Suno’s current platform rules |
Sources, rights and platform boundaries
Make sure you hold the rights to any lyrics, images, trademarks or likenesses you bring into a prompt. Do not imitate a specific living artist, reproduce protected lyrics, or let output imply that a real person or official brand made it.
Availability, pricing, account entitlements and terms for Google products are not the same as the rules of this workspace. When you generate here, this site’s terms, the actual model supply terms and your local law all apply as well.
Google also states that Lyria handles safety through content filters plus recitation checking and artist intent checks. Treat those as model-side guardrails — they do not transfer any rights to you, and they do not replace your own review before publishing.
10 · FAQ
Lyria 3.5 FAQ
What is Lyria 3.5?
Lyria 3.5 is Google DeepMind’s music generation model. It turns text prompts into music audio and lyrics, with Google describing improvements across musicality, lyrics and vocal quality.
How long can a Lyria 3.5 track be?
Google states that tracks can run up to about 3 minutes, and that length is variable — its own examples cover a 60-second clip, a half-length track and a full 3-minute song. Actual output also depends on the prompt and the product surface you use.
Does Lyria 3.5 support Chinese songs?
Google documents multi-vocal generation in eight languages: English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese. Mandarin is not on that documented list, so treat Chinese vocals as something to test in the workspace rather than a documented guarantee.
Can I generate instrumental tracks?
Yes. Google’s guidance is explicit: write "instrumental" in the prompt to exclude vocals completely. Use the negative_prompt parameter as well if you want to push away specific sounds.
Can Lyria 3.5 turn an image into music?
Yes. Google’s model page describes composing with images, and its developer guidance allows up to 10 reference images or PDFs per request. Whether the upload field is available here is shown in the live workspace.
How do I control vocals?
Describe singer demographics and range, voice texture such as gravelly, soulful or breathy, the vocal pattern, and the language. You can also ask for the delivery to change over the track, for example becoming calmer as it progresses.
How do I get a specific song structure?
Either name the sections in the prompt, or use timestamp prompting: Google’s Pro workflows accept [mm:ss] markers, and you can also specify song key, BPM and intensity per segment.
Is Lyria 3.5 free?
Different entry points, regions and accounts follow different rules, and this site does not promise a fixed free allowance. The workspace shows a live point estimate before you submit, and that is the number to check.
Will generated music be marked as AI?
Google states that its Lyria tracks are imperceptibly watermarked with SynthID so that AI-generated or AI-edited audio can be detected.
Is this the official Google website?
No. Nano Banana Pro is an independent multi-model platform and is not the official site of Google, Google DeepMind, Gemini or Flow Music.
Which Google products use Lyria?
Google lists Google Flow Music, the Gemini app, Dream Track in YouTube Shorts, Google Vids, Google AI Studio and the Gemini Enterprise Agent Platform.
Keep exploring AI music
CREATE NEXT
Back to the workspaceTurn a music idea into a draft you can hear
Start from a clear scene, genre or lyric theme, then use the live workspace to refine tempo, instruments, vocals and structure.