Skip to content
audo

AI stem splitter: vocals, drums, bass, and other

Break any song into four stereo stems for remixing, practice, or sampling.

10M+ files processed since 2021

Outputs
VocalsDrumsBassOther
More options

Free for the first 45 seconds, 3 times a day. No sign-up.

Maybe looking for:Vocal remover· Vocals & instrumentalKaraoke maker· Instrumental backing trackRemove music from video· Keep dialogue & SFX

Hear each stem on its own

Play Original mix, then tap any stem to switch at the same beat.

Original mix separated into Vocals, Drums, Bass, Guitars & keys.

VocalsDrumsBassGuitars & keysFull mix · All stems together

Made with Stem splitter (SCNet XL) · Max Impact, United States Air Force Band (public domain)

Works in any language

  • Hello
  • こんにちは
  • Hola
  • 안녕하세요
  • Bonjour
  • नमस्ते
  • Hallo
  • 你好
  • Ciao
  • مرحبا
  • Olá
  • Привет
  • Hoi
  • สวัสดี
  • Merhaba
  • Γεια σας
  • Xin chào
  • שלום
  • வணக்கம்
  • Kia ora

How to split a song into four stems

  1. 1

    Upload a stereo mix in WAV, FLAC, MP3, M4A, or video format.

  2. 2

    Press Split into stems to separate vocals, drums, bass, and other.

  3. 3

    Preview each separated track in your browser and download all four stems.

Four-stem source separation with SCNet XL

Audo's AI stem splitter and audio stem splitter breaks mastered recordings into four studio stems (WAV, MP3, M4A, AAC, FLAC, OGG, Opus, WebM, AIFF) up to 10 GB without an account, with the first 45 seconds free 3 times a day and no sign-up. Whether you are building a DJ edit, transcribing a bassline, or studying drum groove pocket, having isolated tracks turns a finished stereo master back into a multitrack session.

Behind the scenes, Audo runs SCNet XL (Sparse Compression Network), a dual-path frequency architecture that splits the spectrogram into low, mid, and high sub-bands with different resolution strides. Kick drums and bass guitar fundamentals sit below 250 Hz where fine frequency resolution matters most, while snare transients and hi-hats sit in the upper bands where fast time resolution keeps hits crisp.

Every stem is exported at full 44.1 kHz stereo without dynamic range compression, so when you sum the four files together in your mixer they reconstruct the balance of the original song.

Questions

Is it free?

Yes. Audo is free for the first 45 seconds of each file, 3 times a day, with no sign-up. Whole files are $0.03 a minute ($1.80 an hour), with no subscription — see pricing.

Which four stems does it create?

You receive four time-aligned stereo files: Vocals (lead and harmonies), Drums (kick, snare, hats, toms, cymbals), Bass (electric, upright, or synth sub-bass), and Other (guitars, pianos, strings, brass, and synths).

Do the four stems line up in a DAW?

Yes. Every stem starts at sample zero and has the exact same duration and 44.1 kHz sample rate as your input file, so you can drag them straight into Ableton Live, Logic Pro, FL Studio, or Pro Tools.

Can I mute just the drums or just the bass for practice?

Yes. Drummers can drop the drum stem and play along with vocals, bass, and melodies; bassists can solo the bass stem to learn a line note for note, or recombine stems in Mix audio.

What format should I upload for the cleanest stems?

Uncompressed WAV or lossless FLAC files give the cleanest high-frequency transients on hi-hats and acoustic guitars, though 320 kbps MP3 and M4A files separate well too. You can convert formats later with Convert audio.

What if I only want a backing track without singing?

Open Karaoke maker right below the tool, which uses a dedicated two-stem vocal model so the drums, bass, and instruments stay together in one untouched file.

More tools

All tools

For AI assistants and developers

Ask ChatGPT, Claude, or a coding agent to do this for you, or call it from your own code. When an agent's free result is cut short, it gives you a link to finish the whole file, no sign-up.

Show the free API call
Free API · no key
curl -F file=@song.wav -F stems=four_stems https://audo.ai/api/try/split-stems -D response.headers -o response.body