AI Stem Separation
Separate Audio into Individual Stems
Extract vocals, drums, bass, guitar and piano from any song with AI. Get 6 stems plus a ready accompaniment track for remixing, karaoke and production.
- 6 Stems
- ~45 s Processing
- 1 Credit / min
- Operational
- 1-10 cr · Demucs htdemucs 6s · max 10 min
A failed job costs nothing — credits are returned automatically. We keep photo and video results for 30 days, and guest files are deleted after 14 days.
Credit packs from $1.49, one-time. Credits never expire and no subscription is required.
What is Separate Vocals & Instruments?
Separate Vocals & Instruments is an AI tool in Pixelift that runs in the browser with no installation. Extract vocals, drums, bass, guitar and piano from any song with AI. Get 6 stems plus a ready accompaniment track for remixing, karaoke and production. Engine: Demucs htdemucs 6s · max 10 min. Cost: 1-10 credits per result; credits are bought once in packs from $1.49, no subscription, and they never expire. Files are processed on EU servers and never used to train models.
How It Works
- Step 1
Upload Audio
Drop a song or audio file — MP3, WAV, FLAC, OGG, or M4A up to 50MB and 10 minutes.
- Step 2
AI Separates
Demucs splits your audio into 6 stems, and we mix everything except the vocals into an accompaniment track.
- Step 3
Download Stems
Preview each stem individually and download it as a 320 kbps MP3 for your project.
Why Use AI Audio Separation?
6 Stems + Accompaniment
Get vocals, drums, bass, guitar, piano and other instruments, plus the whole song without vocals — each as a separate 320 kbps MP3.
Professional Quality
Demucs (htdemucs 6s, by Meta) gives clean vocals, drums and bass on most studio recordings. Guitar and piano are the hardest stems on dense mixes.
Any Genre
Works with pop, rock, electronic, hip-hop, classical, and more. The AI adapts to different musical styles and arrangements.
Use Cases
Audio separation for creators and producers
Remixing
Import separated stems into your DAW for creative remixes. Adjust levels, add effects, and create entirely new arrangements.
Karaoke
Remove vocals to create karaoke versions of any song. Or isolate vocals for acapella tracks and mashups.
Video Production
Extract background music or isolate dialogue from video soundtracks. Perfect for film editing and content creation.
Music Practice
Isolate specific instruments to practice along with. Mute the drums to play your own, or solo the bass line to learn it.
Frequently Asked Questions
What stems does the AI separate?
Six stems: vocals, drums, bass, guitar, piano and other instruments, plus an accompaniment track — the whole song without vocals.
How long does separation take?
About 30–60 seconds for a typical 3–4 minute song. Recordings can be up to 10 minutes long.
Can I use this for karaoke?
Yes. The vocal removal creates clean instrumental tracks perfect for karaoke. The isolated vocal stem works great for acapella versions.
What audio formats are supported?
Upload MP3, WAV, FLAC, OGG, or M4A files up to 50MB and 10 minutes. Results are delivered as 320 kbps MP3 stems.
Can I import stems into my DAW?
Absolutely. Download the stems and import them into Ableton, Logic Pro, FL Studio, or any DAW for mixing and production.
More tools
Video & audio
Text to Speech
Natural voices in 30 languages from any text.
1-6 kr
- Popular
AI Image Generator
Type what you want to see — get a finished image in seconds.
0.5-7 kr
- Popular
Upscale & Sharpen
A small or soft photo, large and crisp, up to 8×. Usually 17 seconds.
1-4 kr
Colorize Old Photo
Black-and-white photos get natural color. Usually 19 seconds.
1.5 kr
Remove Noise & Grain
Clean up noisy, grainy or over-compressed photos.
0.5-6 kr
Ready to Separate Audio?
Upload a song and get 6 stems plus an accompaniment track in about a minute. Perfect for remixing, karaoke, and music production.
Call it from code
One endpoint, key in the header. The same tool that runs on this page — no browser, no clicking.
curl -X POST https://pixelift.pl/api/v1/generations \
-H "Authorization: Bearer pk_live_..." \
-H "Idempotency-Key: $(uuidgen)" \
-F "tool=audio-separation" \
-F "audio=@nagranie.mp3"Response: { "id": "cmk8…", "tool": "audio-separation", "status": "succeeded", "assets": [{ "kind": "output", "url": "/uploads/…" }], "result": { "stems": ["https://pixelift.pl/uploads/…"], "creditsUsed": 1 } }
A retry with the same Idempotency-Key header is never charged twice; status and the event stream live at /api/v1/generations/{id}. The tool's direct route still works: POST /api/ai-audio/audio-separation
Request fields
tool | string | required | Narzedzie (v1) |
audio | file | required | Utwor do rozdzielenia na slady |