Music engine
An AI music generator for songs, instrumentals and stems
Describe a song and get a finished, produced track back — 30 seconds to 5 minutes, sung or instrumental, split into stems when you want them.
Songwriter mode: one line of text to a finished song
Describe the song, make a few choices, and get back a produced track with arrangement, performance and mix, not a loop to assemble.
- Describe the songMood, instruments, era, subject: up to 2,000 characters of plain English.
- Pick a style, length and tempo97 styles, six lengths from 30 seconds to 5 minutes, and Slow, Medium or Fast.
- Choose who sings, or go instrumentalSet a voice, a language and lyrics, or switch to Instrumental and those controls step aside.
- Press Make the songThe finished track comes straight back, typically in 30 to 90 seconds, and is saved to your library.
In detail
The brief and the result card
- Defaults: Soul, 2 minutes, Medium tempo, female voice, English
- Slow, Medium and Fast mean around 70, 110 and 140 BPM
- A custom style field of up to 120 characters covers anything not on the list
- Instrumental is a real mode: vocals, singing, rap and spoken word are kept out
- Output is MP3 at 44.1 kHz, 128 kbps
- The result card plays in place and shows the style, tempo, voice and language actually used
- Download, Split into stems, Share and Customize artwork sit on the card; click the title to rename
On your phone
- The same engine, styles, languages and prices, laid out for a phone
- A Song / Instrumental switch with starter chips, like a lo-fi study beat or a cinematic trailer score
- Track menu: Recreate with the same settings, Share, Rename, Customize cover, Add to album, Split into stems, Delete
Musician mode: build the band one instrument at a time
Set the tempo, key and chords once, and every instrument is generated as its own stem, told to follow them.
- One master clockTempo, key and time signature are written into every stem's request as exact values, so the stems aim to line up.
- Chords that follow the keyChords picked from the palette or the dice move with you when you change to another major (or minor) key.
- Roll a progressionOne of 12 major or 10 minor music-theory progressions, rolled in your current key.
- Freeze what worksFrozen takes are kept when you regenerate the rest of the band.
- The whole band at onceGenerate every unfrozen instrument in one go; they start in batches of three, about a minute apart.
- Every stem is yoursExport every stem as its own MP3; reopen a production later to restore its key, tempo and length.
Musician mode is marked Experimental: 4/4 is the most reliable time signature, and bar-by-bar chord accuracy lands around 30–60%.
In detail
Master controls
- Tempo: a 60–180 BPM slider with presets at 70, 90, 110, 120, 140 and 160
- Accepted tempo range: 40–220 BPM
- Time signatures: 4/4, 3/4, 6/8, 7/8 and 5/4; odd meters land about 50–70% accurate
- Keys: all twelve roots, C to B, in major or minor
- Length: the same five lengths as Songwriter, applied to every stem
The chord editor
- Tap a bar to swap its chord, add or delete bars up to 8, or reset to Imaj7–vi7–ii7–V7
- Each key's palette: diatonic triads, sevenths and ninths, sus2, sus4 and add9 shapes, and borrowed chords such as ♭VII
- The dice rolls from 22 progressions: 12 major and 10 minor
- Major includes Pop axis, 50s doo-wop, Pachelbel, Jazz turnaround and ii–V–I
- Minor includes Andalusian, Harmonic, Dark pop, Sorrow and Long arc
Working the stage
- Each track has its own description and Generate button; Vocals adds its own lyrics box
- Each instrument is requested solo, and vocals a cappella, so a stem is not a full mix
- Re-roll a take for a fresh one, remove a track, and waiting tracks show Queued
97 styles, 33 vocal languages, and lyrics your way
Pick a named style, choose who sings and in which language, then paste your own words or have them written from one line.
- 97styles in eight categories, plus your own
- 5lead-vocal types, from a female lead to a choir
- 33vocal languages in four groups
- 4,000characters of your own lyrics
- A style, not a guess"Boom-bap" and "Drill" pull different drum programming without you describing either.
- Sung in your languagePick Hindi or Tamil and English vocals are kept out, so the song does not drift back.
- Your exact words, in timeCustom lyrics are synced across the song's sections, so the words land in time.
- A word budget up frontLyrics too long for the length are refused, with the word budget: about 45 words for 30 seconds.
- Lyrics from one lineGive a theme and get singable lyrics in your chosen language, sized to the song's length.
- Edit before it singsWritten lyrics land in the editor, and section labels like "[Chorus]" are never sung.
In detail
All 97 styles
- Moods & feels
- Melancholic, Triumphant, Dreamy, Heroic, Eerie, Foreboding, Romantic, Nostalgic, Hopeful, Dark, Uplifting, Tense, Epic, Cinematic
- Pop & rock
- Pop, Indie pop, Synth-pop, 80s synth-pop, Bedroom pop, Dream pop, Power pop, Rock, Indie rock, Alternative, Classic rock, Punk, Pop-punk, Garage rock, Shoegaze, Post-rock
- Hip-hop & R&B
- Hip-hop, Trap, Lo-fi hip-hop, Boom-bap, Drill, Old-school rap, R&B, Soul, Neo-soul, Funk, Disco, Motown
- Electronic
- House, Deep house, Tech house, Techno, EDM, Trance, Drum & bass, Dubstep, Ambient, Lo-fi, Synthwave, Vaporwave, Chillwave, Trip-hop, IDM
- Acoustic & folk
- Folk, Folk rock, Singer-songwriter, Acoustic, Country, Bluegrass, Americana, Indie folk
- World
- Bollywood, Afrobeat, Latin, Reggae, Reggaeton, Bossa nova, Samba, Salsa, Flamenco, K-pop, J-pop, Indian classical, Arabic, Mediterranean
- Cinematic & score
- Orchestral, Classical, Trailer score, Film score, Minimalist piano, Ambient score, Action score, Romantic score
- Jazz, blues & gospel
- Jazz, Smooth jazz, Bebop, Big band, Swing, Blues, Delta blues, Chicago blues, Gospel, Soul gospel
Voices and languages
- Voices: Female lead, Male lead, Duet, A cappella choir, or Auto to let the engine choose
- Indian (12): Hindi, Punjabi, Tamil, Marathi, Gujarati, Bengali, Telugu, Kannada, Malayalam, Urdu, Odia, Assamese
- English and European (8): English, Spanish, French, German, Italian, Portuguese, Dutch, Russian
- East and South-East Asian (8): Japanese, Korean, Mandarin Chinese, Cantonese, Indonesian, Thai, Vietnamese, Filipino
- Middle East and Africa (5): Arabic, Hebrew, Turkish, Persian, Swahili
- The same styles and languages are on mobile; voice and language hide on instrumentals
Lyric writing
- Written by Claude Haiku 4.5, with the style and mood passed through
- Themes up to 180 characters; under 45 seconds you get a verse and chorus, from three minutes a bridge and outro too
- Your lyrics are stored with the song, so a re-roll sings the same words
Stem splitting: pull a finished track apart
Separate any song you made into instrument stems to preview, download or remix, including a karaoke track without vocals.
- Six stemsVocals, Drums, Bass, Piano / Keys, Guitar, and Other for pads, strings, synths and atmosphere.
- Two stems on your phoneAn a cappella vocal plus the full instrumental, for karaoke backing or a vocal on its own.
- Hear each stem firstPlay any stem in the picker, tick the ones you want, and save each as its own file.
- Mix into one WAVTicked stems are mixed in your browser. Leave out Vocals and you have your own instrumental.
- MP3 or lossless WAVA format toggle in the picker, both at 44.1 kHz.
- Kept with the songOpen a split again later from the result card or the More menu on any song in the gallery.
In detail
What can be split, and the limits
- Only your own music generations, including Musician-mode stems, which can be split again
- Processing typically takes 30 to 90 seconds
- Up to 2 splits per minute; source audio up to 100 MB
- The desktop picker always runs the six-stem split
Your library: covers, albums, a player and share links
Everything you make is kept, given a cover, grouped into albums, playable on every page, and public only if you choose.
- A cover with every songA square, text-free cover made from the song's own title, style, voice, language and lyrics.
- Or design your ownDescribe a cover in up to 500 characters, choose Standard or HD, and generate again until it fits.
- Every song, keptSongs and Productions tabs; play on the card, rename with a click, or delete for good.
- Albums by drag and dropOn desktop, drop one song on another to start an album and drag to reorder, up to 100 tracks.
- A player on every pageA mini player keeps your music playing while you move to other parts of SynergyAI.
- Shared only if you chooseGet a public listening page with a download button; turn sharing off and the link stops working.
In detail
Covers and the gallery
- Covers appear on the result card, gallery, mini player, albums and share cards
- Custom covers show a live preview; Apply to song saves the one you like
- Production cards show key, BPM, time signature and stem count
- Deleting a song removes its stored audio too
Albums and re-rolls
- On a phone, Add to album… offers a new album or any existing one
- Download all saves every track in album order; SynergyAI Files zips an album up to 20 MB and 400 files
- A re-roll makes a new take of the same brief, numbered 1 to 99, and keeps its album position
- Empty albums dissolve on their own
Player and sharing
- Songs and Albums tabs with search, shuffle, and repeat all or one
- Holds up to 400 tracks and 100 albums; hiding it never pauses the music
- Album links list tracks in order, and on mobile one action un-shares a whole album
- Uses your device's share sheet where available, or copies the link
Make music from an agent, Telegram, Slack or email
Ask for a song in a message, or hand the job to an agent, and a finished track comes back.
- Ask in a messageSend a style, length, voice and lyrics on Telegram, Slack or email; the track comes back in the chat.
- Filed where you can find itA song an agent makes is saved to SynergyAI Files, named after the track.
- More than composingAgents can also re-roll, split stems, and build or update albums.
- A real loudness masterMaster one track or a whole album to −14 LUFS with a −1.5 dBTP ceiling, delivered to Files.
In detail
What agents and gateways can do
- Gateway lengths: 30, 60, 120, 180, 240 or 300 seconds
- Gateway prompts are trimmed to 1,200 characters and titles to 100
- Gateway songs are always MP3; lossless WAV is available through agent tools and the API
- Lengths round up: ask an agent for 45 seconds and you get a 60-second song
- The Music Producer agent produces tracks end to end
- More on AI agents and the Workspace
How music is priced
Songs are priced by length, and one account covers every studio: top up as you go, or choose a monthly Pro or Studio plan.
- See the price before you pressEvery length shows its price on the Length picker and the Make the song button.
- Charged only once the song existsA failed song is not charged, and if one cannot be delivered, the charge is refunded.
- Priced by length, not formatWAV costs the same as MP3; a re-roll or a Musician-mode stem costs the same as a song that long.
- No charge to come backRe-opening a split you paid for, building albums and mixing stems in your browser cost nothing extra.
In detail
Covers, lyrics, splits and agents
- Stem splits are priced per started minute of audio, with a one-minute minimum
- The automatic cover costs nothing if the song is yours, finished in the last 30 minutes, and you have had fewer than three no-cost images since
- A custom cover costs the standard or HD image price
- Lyric writing is charged for what it actually uses
- Loudness mastering through an agent costs nothing extra
- Running the Music Producer agent is charged per run
- Your balance is checked before anything is generated
- Composing checks your storage allowance first, before anything is charged
Limits, privacy and rights
What the Music Studio will not do, how often you can use it, and where your rights are set out.
- No audio-to-audioIt cannot remix, cover or extend a recording, and a reference you give an agent is read as text only.
- Per-minute limitsUp to 3 music generations, 2 stem splits and 20 lyric requests per minute, per account.
- A timed-out song starts overA render lost to a timeout cannot be resumed and has to be composed again.
- A private mode that stores nothingIn Forget and Burn, download before you leave. 2-minute cap, MP3 only, no albums, covers, sharing or stems.
- Screened before it startsYour prompt, title and any lyrics pass a child-safety check before the engine is called.
- Rights follow the TermsWhat you may do with a track is governed by the SynergyAI Terms; read them before you release anything.
Frequently asked questions
How does the AI music generator work, and how long does a song take?
Write a plain-English description, pick a style, length and tempo, plus a voice, language and lyrics if it has vocals, then press Make the song. The finished track comes straight back, typically in 30 to 90 seconds, and is saved to your library.
How much does it cost to make a song?
Songs are priced by length; each length's price shows on the Length picker and the Make the song button before you press. WAV costs the same as MP3, and a re-roll the same as a new song. One account covers every studio: top up as you go, or choose a monthly Pro or Studio plan.
Can it sing my own lyrics, and can it write them for me?
Both. Paste up to 4,000 characters of your own lyrics and they are synced so the words land in time; lyrics too long for the length are refused with the word budget, about 45 words for 30 seconds. Or type a one-line idea and Claude Haiku 4.5 writes lyrics you can edit first.
What languages can the vocals be sung in?
33, in four groups: 12 Indian languages including Hindi, Tamil, Bengali and Urdu; 8 English and European; 8 East and South-East Asian, including Japanese, Korean and Mandarin Chinese; and 5 from the Middle East and Africa, including Arabic and Swahili. A non-English choice keeps English vocals out of the song.
Can I split a song into stems, or get a karaoke instrumental?
Yes. A finished track splits into six stems (Vocals, Drums, Bass, Piano / Keys, Guitar and Other) in MP3 or lossless WAV. Tick every stem except Vocals and mix them in your browser for your own instrumental. Your phone also offers a two-stem vocals-plus-music split. Re-opening a split you already paid for is free.
Can I get lossless WAV instead of MP3?
In the studio, songs are composed as MP3 at 44.1 kHz, 128 kbps. Lossless 44.1 kHz WAV is available through agent tools and the API, at the same price, and the stem picker has its own MP3 or WAV toggle. Gateway songs are always MP3.
Is AI-generated music royalty-free, and can I use it commercially?
What you may do with a finished track is governed by the SynergyAI Terms; read them before you publish or license anything. Every track is composed fresh from your description. There is no audio-to-audio path, so it cannot remix, cover or extend an existing recording, and a reference you supply is read as text only.
Can I make music from Telegram, Slack or email?
Yes. Ask on Telegram, Slack or email and the same Music Studio engine sends the track back in the conversation, priced like the studio. Choose 30, 60, 120, 180, 240 or 300 seconds, a style, tempo, voice, language and your own lyrics. Gateway songs are MP3, and prompts are trimmed to 1,200 characters.
Write one line, or build the band
Describe a song and get a finished track with its own cover, or build it one instrument at a time in Musician mode.