How to Make AI Music With Google Gemini and Lyria 3.5

You type “make a birthday song for my sister” into Google Gemini, press Enter, and get a short generic pop track that could have been written for almost anyone.

It’s tempting to blame the AI when that happens. But with Google’s latest music-generation model, the quality of the result often depends on two things people overlook: how you write the prompt and which Gemini model you have selected.

Google made Lyria 3.5 available to Gemini users on September 4, 2026, bringing AI music generation to both the web and mobile versions of Gemini. The model follows earlier Lyria releases from this year, adding more expressive vocals, stronger song structures, and ready-made templates.

The good news is that you don’t need to be a musician or know music production to use it.

In this guide, I’ll show you how to make a song with Google Gemini, how to create longer tracks, how to write better Lyria 3.5 prompts, and how to download and share your AI-generated music.

What Is Google Gemini’s AI Music Generator?

Google’s Gemini app can generate original music using its Lyria family of music-generation models.

Lyria 3.5 is designed to turn natural-language instructions into complete musical tracks. You can describe the genre, instruments, tempo, vocals, mood, lyrics, and overall structure you want, and Gemini generates a song based on those instructions.

You can create music for things such as:

• Birthday songs
• Personal gifts
• YouTube videos
• Podcasts
• Social media content
• Short films
• Travel videos
• Background music
• Gaming videos
• Instrumental tracks
• Personal experiments

You can also provide an image or video and ask Gemini to create music that matches the content.

Google’s music-generation system has evolved quickly in 2026. Lyria 3 initially introduced shorter 30-second generations, while Lyria 3 Pro subsequently expanded the available track length for eligible paid users. Lyria 3.5 builds on that foundation with more expressive vocals and additional controls in the Gemini app.

What You Need to Create Music in Gemini

Before looking for the music-generation option, make sure your Gemini account meets Google’s requirements.

You Must Be 18 or Older

Google currently requires users to be at least 18 years old to access the music-generation feature.

You Need a Supported Google Account

You can use a personal Google Account or, where supported, an eligible work or school account.

Availability can depend on the type of account and Google’s regional rollout.

Keep Activity Must Be Enabled

This is the requirement that can easily be missed.

Gemini’s Keep Activity setting must be turned on for the music-generation feature to work.

If you previously disabled Keep Activity because you didn’t want Gemini storing your activity, the Create music option may not appear.

That can make it look like your account doesn’t have access to Gemini’s music generator when the setting is actually preventing the feature from working.

If you can’t find the music option, check this setting before troubleshooting anything else.

Lyria 3.5 Availability

Google says Lyria 3.5 is available in countries where the Gemini app itself is available, although individual features can still vary according to account type, subscription, age, and rollout status.

How to Create a Song in Google Gemini

The basic process is nearly identical whether you’re using Gemini on a computer or the mobile application.

Step 1: Open Gemini

Go to Gemini in your browser or open the Gemini app on Android or iPhone.

Start a new conversation.

Step 2: Open the Tools Menu

Select Upload & tools, represented by the + icon.

Look for Create music in the available tools.

Step 3: Choose a Music Template

Gemini can provide templates that make it easier to get started.

Choose one if you want a quick starting point, or skip the templates and describe exactly what you want in your own prompt.

Step 4: Select Vocals or Instrumental

Decide whether your track should contain singing.

You can choose:

• Vocals
• Instrumental

If you select vocals, your prompt can also describe the type of voice and singing style you’re looking for.

Step 5: Select the Track Length

Depending on the model and account configuration, Gemini provides short and full-length generation options.

A short track is roughly one minute, while a full generation can run approximately two to three minutes.

Step 6: Add an Image or Video if You Want

Gemini can also use an uploaded photo or video as inspiration for the soundtrack.

For example, you could upload footage from a hiking trip and ask for an atmospheric cinematic instrumental track.

Step 7: Write Your Prompt

Describe the song you want as specifically as possible.

Then send your request and wait for Gemini to generate the track.

Once it’s finished, Gemini presents the music directly in the conversation.

How to Make Longer Songs in Gemini

One of the most important settings to check is the Gemini model selector.

According to Google’s documentation, creating a full-length track requires selecting Pro in the model picker.

If Gemini remains on the default model, you may receive the shorter version even if your prompt asks for a long song.

This is one of the easiest things to overlook when experimenting with the music generator.

Free vs Paid Gemini Music Generation

Lyria 3.5 is available to Gemini users, but Google applies usage limits depending on the account and plan.

Google has not publicly documented a single universal generation quota that applies to every Gemini plan.

Earlier reporting around the launch of Lyria 3 Pro indicated different daily limits for Google AI Plus, AI Pro, and AI Ultra subscribers, but those numbers should not be assumed to represent the current Lyria 3.5 limits.

The practical approach is simple: Gemini will notify you when you’ve reached your available generation limit.

For experimentation, start with shorter tracks. Once you’ve developed a prompt that produces the sound you want, switch to the appropriate Pro model and generate the longer version.

That can help you avoid wasting your available generations on full-length tracks that need major revisions.

How to Write Better Gemini Music Prompts

The biggest mistake beginners make is writing prompts that are too vague.

Lyria can generate music from a simple sentence, but detailed instructions give the model much more information about what you’re trying to create.

Google’s guidance emphasizes several useful elements.

1. Specify the Genre

Tell Gemini what kind of music you want.

Examples:

• 1990s Bollywood romance
• Modern indie folk
• Lo-fi hip-hop
• Acoustic pop
• Cinematic orchestral
• 1980s synth-pop
• Melodic house
• Jazz piano
• Emotional R&B

You can combine genres when it makes sense.

2. Describe the Tempo and Rhythm

Don’t simply say that the song should be “good” or “catchy.”

Describe its movement.

For example:

“A slow, relaxed rhythm with a gentle swaying groove.”

Or:

“An energetic 140 BPM dance beat with a driving rhythm.”

3. Name the Instruments

Specific instruments give Gemini a clearer musical direction.

You might request:

• Acoustic guitar
• Piano
• Electric guitar
• Tabla
• Violin
• Synthesizers
• Saxophone
• Hand percussion
• String section
• Drum machine

4. Describe the Vocals

If you’re generating a vocal track, explain what you want the singer to sound like.

For example:

“Warm female vocals with an intimate, breathy delivery.”

Or:

“Powerful male vocals with layered choir harmonies during the chorus.”

5. Explain the Story

This is particularly important when you’re creating a personalized song.

Instead of:

“Make a birthday song for my sister.”

Give Gemini actual information.

For example:

“Write an upbeat acoustic pop song for my sister Priya’s 25th birthday. She loves chai, late-night drives, and her cat Mochi. Use warm female vocals, handclaps, and a memorable chorus that repeats her name.”

The second prompt gives the model much more material to work with.

It knows who the song is for, why you’re making it, what details to mention, what instruments to use, what the vocals should sound like, and what the chorus should do.

That’s what makes the resulting song feel personalized.

Use Action Words in Your Music Prompts

Google recommends beginning prompts with an action such as:

• “Write…”
• “Compose…”
• “Create…”

This makes your intention clearer.

You can also tell Gemini how the song should develop.

For example:

“Compose a cinematic instrumental track. Open with a quiet 10-second piano introduction, gradually introduce strings, then build toward a powerful final section.”

This gives Lyria information about both the sound and the structure.

Tell Gemini How You Want the Song Structured

You don’t have to describe the entire song in musical terminology.

Simple structural instructions can help:

• Intro
• Verse
• Pre-chorus
• Chorus
• Bridge
• Final chorus
• Outro

For example:

“Start with an eight-second acoustic guitar intro, move into a verse, build through a short pre-chorus, then finish with a large, catchy chorus.”

Lyria 3.5 is designed to understand these kinds of musical instructions.

Can You Provide Your Own Lyrics?

Yes. If you’ve already written lyrics, you can include them in your prompt.

For more structured generation, Google’s Lyria API documentation also describes section markers such as:

[Verse]

and

[Chorus]

However, Google doesn’t necessarily document every such control as a guaranteed feature of the consumer Gemini interface.

Therefore, treat these markers as useful formatting rather than a promise that the Gemini app will always interpret every tag exactly as intended.

A Simple Formula for Better Gemini Music Prompts

If you’re not sure what to write, use this formula:

Action + genre + era + mood + tempo + instruments + vocals + subject + structure

For example:

“Create a warm acoustic pop song inspired by 2000s romantic music, with a relaxed mid-tempo rhythm, acoustic guitar and soft piano, featuring warm female vocals. The lyrics should celebrate my sister Priya’s 25th birthday and mention chai, late-night drives and her cat Mochi. Start with a short guitar intro and build into a catchy chorus that repeats her name.”

This gives Gemini far more direction than simply asking it to “make a birthday song.”

How to Create a Soundtrack From a Photo or Video

Gemini can also generate music based on visual content.

Upload a photo or video and describe the soundtrack you want.

For example:

“Create an instrumental soundtrack for this hiking video. Make it cinematic and atmospheric, starting quietly and gradually building toward an uplifting ending.”

Gemini analyzes the visual material and uses it as additional context for the music.

A sunset over the ocean might inspire a calm, atmospheric composition, while a crowded festival video could lead to something more energetic.

However, don’t leave everything to the AI’s interpretation.

Tell Gemini what you want.

The uploaded media gives it visual context, while your text prompt tells it how that context should be translated into music.

How to Download a Song Created in Gemini

After Gemini finishes generating your track, you can listen to it directly inside the conversation.

Depending on the available options, you can download the generated content in different formats.

MP3

The MP3 version contains the audio itself.

It’s useful if you want to:

• Edit the music in another application
• Use it as a podcast intro
• Add it to a video
• Create a personal ringtone
• Store the track as an audio file

MP4

The MP4 version combines the audio with generated cover artwork.

This can be useful when you want to share the song on platforms that expect a video file.

Gemini can generate cover art using Google’s Nano Banana image-generation technology.

You can also share the generated track using a shareable link when the feature is available on your account.

Gemini AI Music Includes SynthID

AI-generated tracks from Google’s systems can include SynthID, Google’s technology for identifying AI-generated media.

SynthID uses an inaudible watermark designed to remain detectable even after certain modifications to the generated content.

Google also provides ways to check media for its AI watermark. For example, you can provide supported audio content to Gemini and ask whether it contains a Google AI watermark.

This is part of Google’s broader effort to make AI-generated content more identifiable.

What Gemini’s AI Music Generator Can’t Do

The technology is impressive, but there are still important limitations.

You Can’t Simply Clone a Famous Artist

Don’t expect Gemini to reproduce the exact voice of a specific singer.

Requests involving the direct imitation of a particular artist’s voice may be restricted.

Instead of asking for:

“Make this sound exactly like [artist].”

Describe the musical characteristics you want.

For example:

“Create energetic 2000s Punjabi pop with bright percussion, catchy melodies and expressive male vocals.”

That gives Gemini a musical direction without asking it to reproduce a specific artist.

Don’t Expect Precise Track Editing

Another important limitation is that music generation isn’t the same thing as editing an existing multitrack project.

If your first generation has a chorus that’s too quiet, asking Gemini to make only that chorus louder may produce a new generation rather than surgically modifying the original audio.

For major changes, rewrite the prompt and include the desired adjustment from the beginning.

For example:

“Create the same general concept, but make the chorus more powerful, increase the vocal presence, and add layered harmonies.”

You’ll still get a new generation rather than a traditional audio-editing operation.

Every Generation Can Be Different

AI music generation isn’t perfectly deterministic from a user’s perspective.

The same prompt can produce different musical results.

That’s actually useful.

If the first version isn’t right, generate another. You may get a better melody, vocal performance, arrangement, or chorus on the next attempt.

For an important project, generating several versions and choosing the strongest one can be more effective than endlessly trying to perfect a single generation.

Can You Use Gemini-Generated Music Commercially?

This is an area where creators should be careful.

Whether you can commercially use a particular AI-generated song depends on Google’s current terms, the applicable Gemini plan and feature conditions, and the laws governing copyright and AI-generated works.

AI copyright law also remains unsettled in several jurisdictions.

There have already been significant legal disputes involving AI-generated music and copyrighted works. For example, a German court ruling involving AI music company Suno and collecting society GEMA received attention in 2026.

That doesn’t automatically determine what you can or cannot do with a song generated through Gemini.

If you’re planning to use an AI-generated track in:

• A monetized YouTube channel
• A commercial advertisement
• A client project
• A paid podcast
• A video game
• A business campaign

read the applicable Google terms and licensing conditions before publishing it commercially.

Tips for Getting Better Results From Gemini Music

If your first songs sound generic, don’t immediately assume the model isn’t good enough.

Try improving the prompt.

Start With a Short Generation

Use a short track to test your concept.

Listen for:

• Vocal style
• Melody
• Tempo
• Instrumentation
• Overall mood
• Song structure

If something is wrong, change the prompt.

Once the direction works, generate the longer version.

Add Concrete Details

“Happy song” is vague.

“Upbeat acoustic pop with handclaps, warm female vocals, a 110 BPM groove and a bright guitar hook” is much more specific.

Give the Song a Story

A song with a subject, setting, people, and specific details is more likely to feel distinctive.

Instead of:

“Make a romantic song.”

Try:

“Write a nostalgic acoustic song about two people meeting again after ten years in their hometown.”

Don’t Overload the Prompt

More detail isn’t always better.

Give Gemini the most important information first and avoid stuffing the prompt with contradictory instructions.

Generate Multiple Versions

If the first version is close but not perfect, try again.

Small changes to the prompt can produce surprisingly different arrangements.

Gemini Music vs. Traditional Music Production

AI music generation is useful when you want to turn an idea into a complete track quickly.

Traditional music production still offers substantially more control over individual elements.

With a conventional DAW, you can independently manipulate:

• Individual instruments
• Vocals
• Effects
• Volume automation
• Equalization
• Compression
• Timing
• MIDI notes
• Arrangement
• Mixing
• Mastering

Gemini’s music generator is instead focused on turning a natural-language concept into a finished musical result.

For casual creators, content creators, and people who simply want to experiment with music, that simplicity is one of its biggest advantages.

Final Thoughts

Creating a song with Google Gemini is remarkably simple. The harder part is getting the AI to create something that actually feels personal and intentional.

Lyria 3.5 gives Gemini users a more capable music-generation system, but the quality of the result still depends heavily on the instructions you provide.

If the song sounds generic, don’t just blame the AI. Start by checking that Keep Activity is enabled, make sure you’re using the appropriate Pro model for full-length generation, and give Gemini concrete information about the genre, tempo, instruments, vocals, subject, and song structure.

Start with a short generation, refine your prompt, and only then spend a full-length generation on the version you actually want.

The difference between “make a birthday song for my sister” and a detailed prompt filled with names, memories, instruments, and musical direction can be enormous.

With the right prompt, Gemini’s AI music generator can go from producing a generic 30-second tune to creating something that actually feels like your song.