Skip to content
Aleatoria

AI music for DJs and producers

A track you can actually play on Friday.

Describe it in one sentence — the synthesiser or the AI model composes it, renders it in the EU, and you take it with you.

Opens the studio with your prompt. The account is free.

Open the studio

Hear what comes back

One procedural track in three takes, one song from the AI model and one nature scene. It plays right here; nothing downloads until you tap.

  • Take A

    Instrumental · Synthesizer

    dark hypnotic techno, rolling bassline, warehouse reverb

    46.3 s124.0 beats per minuteAm

  • Take B

    Instrumental · Synthesizer

    dark hypnotic techno, rolling bassline, warehouse reverb

    46.3 s124.0 beats per minuteAm

  • Take C

    Instrumental · Synthesizer

    dark hypnotic techno, rolling bassline, warehouse reverb

    46.3 s124.0 beats per minuteAm

  • Sung demo, neural engine

    Song · ACE-Step 1.5

    melancholic house with a female vocal, late night city

    45.0 s

    0:00 / 0:45

  • Demo: night rain on a tin roof

    Nature · Synthesizer

    night rain on a tin roof with distant thunder and crickets

    45.0 s

    0:00 / 0:45

Three takes, one request

Same prompt, same tempo, same key. Three times in a row. Switch A/B/C while it plays — the playhead stays where it is. You could mix them into each other, but it is different music.

locked: tempo 124.0 BPM, key A minor, mutation off, one attempt

plays right here in the browser · 1.9 MB MP3 320 kb/s · nothing downloads until you click

124.0 beats per minuteAmseed 15109831720:00 / 0:46

124.0 beats per minuteAmseed 5917361470:00 / 0:46

124.0 beats per minuteAmseed 15284924360:00 / 0:46

Selected: take A
The numbers, and how we measure them

Waveform cross-correlation: A↔B correlation -0.007A↔C correlation -0.009B↔C correlation -0.053

Measured on these three files, 10 September 2026.

how we measure it: Waveform cross-correlation between pairs of the final renders, computed on the exact files playing here. Nothing else feeds that number.

Any two unrelated recordings also come out near zero — on its own the number proves nothing. So listen. It only confirms what you hear.

The previews on this page are 320 kb/s MP3. The originals are 24-bit WAV 48 kHz.

Start from a genre

Starting points, not finished tracks: a tap opens the studio with this prompt for you to edit.

  • Technodark hypnotic techno, rolling bassline, 126 BPMOpen in the studio
  • Deep housedeep house, warm chords, 122 BPMOpen in the studio
  • Drum & bassliquid drum & bass, 174 BPMOpen in the studio
  • Tranceuplifting trance, 138 BPMOpen in the studio
  • Lo-fi hip hoplo-fi hip hop for studying, dusty drums, 84 BPMOpen in the studio
  • Ambientbeatless ambient, slow padsOpen in the studio
  • Cinematiccinematic score, strings and tensionOpen in the studio
  • Popfresh pop, synths, 110 BPMOpen in the studio
  • Jazzlate-night jazz, brushes and upright bassOpen in the studio
  • Folk / acousticacoustic folk, guitar, 96 BPMOpen in the studio

The DJ pack and a sung song

DJ pack: beatgrid and cue points in the MP3 and FLAC

The MP3 and the FLAC carry a locked beatgrid, hot cues by section and a loop on the drop — Serato reads them on import, and for rekordbox the pack holds an XML. You download the pack from the studio after generating.

AI model: a song with vocals

ACE-Step 1.5 composes a song with a real vocal and lyrics — up to 4 minutes for 5 credits.

Measured, not promised

Three figures from the very MP3 you play and download here — not from a master you do not have at hand.

true peak

-0.97 decibels full scale

RMS

-16.15 decibels full scale

clipped samples

0

Measured on the 320 kb/s MP3 playing here — on take A, not on some other file. Download it and check for yourself.

Masters are rendered to a true peak of -1.00 dBFS. A lossy codec does not hold that exactly through a decode, so the measured peak sits a little off it — by hundredths of a decibel at 320 kb/s. We quote it from the MP3 you have, not from a WAV you do not.

measured 21 September 2026

For DJs: a file you can mix

Whether it mixes is decided by a folder, not an adjective. This is what one request sends back:

aleatoria_techno_124_Am.mp3

Instrumental · Synthesizer

AI-GENERATED
BPM
124.0 beats per minute
key
Am (8A)
length
46.3 s
format
MP3 320 kb/s · original 24-bit WAV 48 kHz
true peak
-0.97 decibels full scale
seed
1510983172

8A is our own Camelot conversion. The file’s tag says Am. The AI Act Art. 50 marking is written into the file itself, in the ID3 tag — not only here on the page.

The finished track’s folder

tracks/5477fe93-926d-4b24-9aa7-c7a968b3a548/
  mix.wav      ← 24-bit / 48 kHz · mastered to true peak -1.00 dBFS
  mix.flac     ← lossless, 24-bit
  mix.mp3      ← 320 kb/s preview
  score.json   ← the whole score: chords, sections, seed
  score.mid    ← SMF type 1
  stems/kick.wav  snare.wav  hats.wav  bass.wav  pad.wav  lead.wav
  stems/*.flac

SHA-256 of the take A MP3 you can download here — check that what you heard is what lands in your folder.

shortened SHA-256 digest of the downloadable file

Download this render 1.9 MB · MP3 320 kb/s · no account, no email

The pack with the WAV, the stems and the MIDI comes out of the studio, after you generate. When there is one to download, there will be a link here — not before.

the request

prompt
dark hypnotic techno, rolling bassline, warehouse reverb
tempo
124.0 beats per minute
key
A minor
length
46.3 s
engine
Synthesizer

You set the tempo and the key. That is exactly what comes out: 124.0 BPM asked for, 124.0 BPM in the tag.

Not convinced? Download the MP3 and open the tags: TBPM 124, TKEY Am.

The stems sum back to the mix. Nothing goes missing on the way.

Rekordbox, Serato and Traktor read 124.0 BPM and Am straight from the MP3 and FLAC tags on import.

44.1 / 48 / 96 kHz, WAV and FLAC, 24 bit. Downloads are unlimited on every plan, including Free.

The stems follow the arrangement — what is not in it has no file of its own.

This is for procedural tracks. Tracks from the AI model have no MIDI of their own (no score; on Studio and Pro the AI stems tool transcribes one) and their tags carry the BPM only when we measured a steady beat, never the key.

  • 24-bit WAV / FLAC · 44.1 / 48 / 96 kHz
  • stems + MIDI
  • seed = byte-identical render
  • hosted in the EU

Two engines

tier 0 — procedural

Our own composer and our own synthesis: Markov chord progressions, motif development, per-genre arrangement, subtractive polyBLEP, FM, Karplus–Strong, additive pads, noise textures, FDN reverb, sidechain. Gives you stems, MIDI and a seed.

0 credits. Always. On every plan.

tier 1 — neural

ACE-Step 1.5 for sung songs with real vocals and lyrics. It is not ours — it is an open model. No MIDI of its own: there is no score behind it; on Studio and Pro the AI stems tool transcribes one from the stems.

5 credits per song, up to 4 minutes.

nature — synthesiser plus an AI texture

14 kinds of nature. Not a single sample.

All of it computed, none of it played back from a library — so there is no loop point and no two renders alike.

These 14 are procedural and cost 0 credits. The studio also has an AI texture engine (stable_audio) — that one costs 2 credits per clip.

More about the engines

Where it falls short today

The neural engine is ACE-Step 1.5 — an open model we did not train. Our own model is a plan, not this release.

Suno has more users and a longer history. We have stems, MIDI, seeds, a tempo and key you can lock, and EU hosting.

Working today

The studio, both engines, the editor, exports, downloads.

Not yet

Registration, and therefore generating on an account.

The comparison, straight

criterionAleatoriaSuno
downloadsunlimited on every plan, including FreeFree 7 total, Pro 20 a month
stemsone file per instrument, they sum back to the mixwe did not verify this
MIDISMF type 1 for every procedural trackwe did not verify this
seedbyte-identical regenerationwe did not verify this
BPM and key in the tagsyes, on procedural trackswe did not verify this
tempo and key you can lockyou set them in the request, that is what comes outwe did not verify this
procedural enginefree on every planwe did not verify this
hosting and GDPRdata and audio in the EU (Google Cloud, europe-west4); account e-mails are sent by Resend (USA)we did not verify this

  • downloads

    Aleatoria unlimited on every plan, including Free

    Suno Free 7 total, Pro 20 a month

  • stems

    Aleatoria one file per instrument, they sum back to the mix

    Suno we did not verify this

  • MIDI

    Aleatoria SMF type 1 for every procedural track

    Suno we did not verify this

  • seed

    Aleatoria byte-identical regeneration

    Suno we did not verify this

  • BPM and key in the tags

    Aleatoria yes, on procedural tracks

    Suno we did not verify this

  • tempo and key you can lock

    Aleatoria you set them in the request, that is what comes out

    Suno we did not verify this

  • procedural engine

    Aleatoria free on every plan

    Suno we did not verify this

  • hosting and GDPR

    Aleatoria data and audio in the EU (Google Cloud, europe-west4); account e-mails are sent by Resend (USA)

    Suno we did not verify this

An empty cell means we did not verify it — not that they cannot do it. Verified on suno.com 12 October 2026. Better still, check it yourself.

What it costs

  • Free

    listening, the library and a 30 s preview without an account

  • Basic

    €2.99 a month

    • 100 credits a month
    • about 20 songs from the AI model
    • commercial rights

  • Studio

    €9.99 a month

    • 500 credits a month
    • about 100 songs from the AI model
    • commercial rights

  • Pro

    €29.99 a month

    • 1,500 credits a month
    • about 300 songs from the AI model
    • commercial rights

The procedural engine costs 0 credits on every plan. Top-up packs never expire.

Prices include Czech VAT at 21% — the same wherever you are in the EU.

Full pricing and the Suno comparison

Your first track. Tonight.

Describe it in one sentence, hear what comes back, and keep the one you will play.

Open the studio

Common questions

Is it free?

A 30 s preview plays without an account on the Try page. The account is free — listening and the library. Rendering whole tracks is on the Basic plan and up. Basic starts at €2.99 a month. Try 30 s without an account

What do I get to download?

24-bit WAV, FLAC and 320 kb/s MP3; procedural tracks also come with stems per instrument and MIDI. Downloads are unlimited on every plan.

Where does it run, and who sells it?

Everything runs in the EU — Google Cloud, europe-west4. Seller: Cloudpeakify s.r.o., Prague. Prices include Czech VAT at 21% — the same wherever you are in the EU. Legal documents