AI Music Generation · Text to Music

Describe the track. Get instrumental or vocal music, up to 3 minutes.

Three real Google Lyria models cover everything from a quick 30-second instrumental bed to a full ~3-minute song with structure, vocals and timed lyrics. Prompt helper tags insert timestamp, BPM, key and intensity syntax without memorizing it, and a Fixed-seed option makes a good take reproducible instead of a one-shot roll of the dice.

Generate music → See all 3 models →

1 Models 2 Prompt Helpers 3 Vocals & Lyrics 4 History 5 Who It's For

Lyria models

~3 min

max length, Lyria 3 Pro

prompt helper tags

vocal modes

What is SparkStation's AI Music Generation? A prompt-driven music generator built on 3 real Google Lyria models, Lyria 2 (30 seconds, instrumental-only, negative-prompt support), Lyria 3 Clip (30 seconds, optional vocals) and Lyria 3 Pro (up to ~3 minutes, full song structure, vocals, timed lyrics). Describe genre, mood, instruments, BPM and key, or use insertable helper tags for timestamp, BPM, key and intensity, toggle Instrumental or With Vocals, and optionally lock a seed for a reproducible result.

How it works

One prompt. Instrumental or vocal, 30 seconds or a full song.

No separate composer tool and lyricist tool, one form covers model, prompt, vocals and reproducibility.

Choose a model

Lyria 2 for a quick instrumental, Lyria 3 Clip for a short vocal-capable clip, or Lyria 3 Pro for a full ~3-minute song.

Write the prompt

Genre, mood, instruments, BPM, key, typed directly or inserted via helper tags.

Set vocal mode

Instrumental, or With Vocals with your own lyrics and vocal style.

Generate & manage

Every track saves to History, playable, downloadable, with lyrics files where relevant.

🧠 Step 1 · Choose

Three Real Lyria Models, Each With a Different Job"Choose Model" modal, Google · Lyria, confirmed live in the product

A 30-second instrumental sting and a full song with verses and a chorus aren't the same job, so SparkStation exposes three distinct models rather than one generic "make music" button.

Lyria 2

30sNegative promptInstrumentalSeed

Instrumental-only, with support for a negative prompt to steer away from unwanted elements, the quickest option for a short background bed.

Lyria 3 Clip

V3 · ⚡1

30sVocalsSeed

The same 30-second length as Lyria 2, but with optional vocals, a fast way to test a vocal idea before committing to a full-length generation.

Lyria 3 Pro

PRO · ⚡2 · Default

Up to ~3 minVocalsSeed

The default selection. Full song structure, vocals and timed lyrics, read verbatim from the product's own description: "Up to ~3 min · full song structure · vocals · timed lyrics."

Try a model →

✍️ Step 2 · Describe

Prompt Like a Composer, Not a Guesser5 insertable helper tags, real examples from the product's own history

The Prompt field takes plain description, genre, mood, instruments, BPM, key, and five helper chips insert the right syntax without memorizing it: +Timestamp, +BPM, +Key, +Intensity, +Instrumental. That timestamp support matters beyond convenience: it's what makes scene-targeted scoring possible in the first place.

+ Timestamp + BPM + Key + Intensity + Instrumental

Real example prompt (Lyria 3 Pro, instrumental)

[0:00-0:20]: intro music, bass at lowfi and a mix of violin and flute

Real example prompt (Lyria 3 Pro, instrumental)

travelling bgm [0:00-0:30]

Write a prompt →

🎤 Step 3 · Add vocals

Instrumental by Default, Vocals and Lyrics When You Want ThemVocal Mode toggle + Vocal instructions / lyrics field, verified live

A two-way Vocal Mode toggle, Instrumental / With vocals, sits on every vocal-capable model. Switching to With Vocals reveals a second field, Vocal instructions / lyrics, for verse and chorus text, a vocal style direction, and even a language, all folded into the prompt at generation. The product's own helper text states this plainly: "Folded into the prompt at generation. Tip: you can specify the language here too (e.g. 'in Spanish')."

On Lyria 3 Pro specifically, vocal tracks come with timed lyrics, confirmed by real generated tracks in the product's history that show a separate downloadable "Lyrics" file alongside the audio, not just an internal generation detail.

Instrumental With vocals

Vocal instructions / lyrics

e.g. verse & chorus lyrics, vocal style, language…

Real example prompt (Lyria 3 Pro, with vocals)

Country genre travel song with philosophy

Add vocals →

🗂️ Step 4 · Manage

Every Track, With Its Prompt AttachedGeneration History, List and Grid views, Refresh

Past generations live in a Generation History panel, switchable between List and Grid, each entry shows the prompt used, tag chips for instrumental/vocals, duration and model, with play, download, edit and delete actions, plus a separate Lyrics download for vocal tracks.

♪

Ravanan Test

"0:00-00:20 : intro music, bass at lowfi and a mix of violin and flute"

instrumental1:44Lyria 3 Pro

EditDelete

♪

My Way

"Country genre travel song with philosophy"

vocals2:59Lyria 3 ProLyrics file

EditDelete

View music history →

Why it works better

Most AI music tools stop at "make me a song." This one scores a scene.

A generic single-model music generator takes a mood word and returns a fixed-length loop. Scoring an actual video means hitting specific moments, not just matching a vibe.

✕ Generic single-model music generator

One model, one fixed length, no room between a 30-second loop and a full song
No timestamp syntax, so cueing a change at a specific moment means trimming after the fact
Vocals, if supported at all, come with no lyrics control and no timed-lyrics file
No seed control, every regeneration is a fresh roll, even for a take you liked

✓ SparkStation AI Music Generation

3 real Lyria models spanning a 30-second clip to a ~3-minute full song structure
Timestamp, BPM, key and intensity helper tags for scene-targeted scoring
A dedicated Vocal instructions/lyrics field, with a downloadable timed-lyrics file on Lyria 3 Pro
A Fixed-seed option to reproduce a take instead of gambling on a new one

Who it's for

Anyone scoring a video, not just looking for a loop.

🎬

Filmmakers & editors

Timestamp-targeted scoring for a specific cut, synced to picture rather than trimmed after the fact.

📱

Social & video creators

Background music for a vlog, reel or travel edit, generated to match a mood and duration in one pass.

📣

Marketers & ad teams

A jingle or an on-brand instrumental bed for a promo, with a negative prompt to keep unwanted tones out.

🎙️

Podcasters & game devs

Short intro/outro stings or ambient loops, reproducible via a fixed seed once the right take is found.

Works well with

Music rarely stands alone in a finished video.

Answers

AI Music Generation, in brief.

What is SparkStation's AI Music Generation?

A prompt-driven music generator built on 3 real Google Lyria models, Lyria 2, Lyria 3 Clip and Lyria 3 Pro. Describe genre, mood, instruments, BPM and key, optionally add vocals with your own lyrics, and generate an instrumental or vocal track.

Which music models are available?

Three, all under Google · Lyria: Lyria 2 (30 seconds, instrumental only, supports a negative prompt), Lyria 3 Clip (30 seconds, optional vocals), and Lyria 3 Pro (up to roughly 3 minutes, full song structure, vocals, timed lyrics, the default selection).

Can I generate vocal tracks with lyrics?

Yes, on Lyria 3 Clip or Lyria 3 Pro. Toggling Vocal Mode to "With vocals" reveals a Vocal instructions / lyrics field, verse and chorus lyrics, vocal style, even the language to sing in, which is folded into the prompt at generation. Lyria 3 Pro additionally produces timed lyrics, downloadable as a separate file.

Can I score a specific moment in a video?

Yes. The Prompt field accepts timestamped instructions directly, e.g. "[0:00-0:20]: intro music, bass at lowfi and a mix of violin and flute", and a +Timestamp helper tag makes this easy to insert without typing the bracket syntax by hand.

How long can a generated track be?

Up to roughly 3 minutes on Lyria 3 Pro, which supports full song structure. Lyria 2 and Lyria 3 Clip are shorter, 30-second clips.

Can I get the same track again if I like it?

Yes, an Output Mode setting lets you set a Fixed seed (any number you choose) instead of leaving it blank for a random result each time, useful for regenerating small variations without losing the overall take you liked.

Is there a free way to try it?

Yes, AI Music Generation has its own free-to-start entry point alongside every other SparkStation tool.