免费试用
10 分钟
≈ 3 首歌
无需信用卡。免费分钟会一直保留到你真正使用为止。
Powered by htdemucs
AI Stem Splitter & Vocal Remover: Karaoke, Remix, DJ Acapellas
The AI vocal remover and 6-stem splitter for karaoke tracks, DJ acapellas, remix samples, and guitar practice loops — any song, under 60 seconds. Start free with 10 minutes of processing. No subscription, credits never expire.
Start your first free split🎤 Vocals🥁 Drums🎻 Bass🎸 Guitar🎹 Piano🎵 Other
把音频文件拖到这里
或点击浏览文件
MP3 · WAV · FLAC · M4A最大 50 MB
Demo 预览 · 真实的分离现场
选一个风格，切 6 轨 / 4 轨，点任意 stem。
00:00 / 00:20
Core stems come from htdemucs_ft. Guitar and piano come from the 6-stem model htdemucs_6s, marked with the 6S badge.
CC-BY · “Traveling Lights” by Zep Hurme ft. DeBenedictis · source
Feature breakdown
Everything you need for vocal removal and stem separation
Professional-grade AI vocal removal and stem separation. Remove vocals, create instrumentals, and split stems — all in one tool.
- Up to 6 stems
AI-Powered 6-Stem Separation
Split any song into six clean tracks — vocals, drums, bass, guitar, piano, and other — from a single upload. One click, no plugins, no DAW required.
- Meta AI research
Powered by htdemucs — Meta AI's Sony MDX Winner
Every upload runs on htdemucs, the 2022 hybrid-transformer model from Meta AI (Rouard et al.) that won the Sony Music Demixing Challenge. The same model top studios and researchers use — now one click away.
- One click
AI Vocal Remover & Karaoke Maker
Remove vocals from any song to create instant karaoke tracks, or isolate acapellas for remixing. Clean separation with minimal bleed — perfect for singers, DJs, and content creators.
- Direct processing
YouTube & SoundCloud Vocal Remover
Process videos and tracks directly from their URLs — just paste a YouTube or SoundCloud link. No downloading, no converting, we handle everything automatically.
- Jam-ready
Built for Practice & Production
Mute the drums to jam with a backing track, solo the bass to transcribe a line, or loop a guitar phrase to learn it bar by bar. Every stem is a new way to play, learn, or create.
- Free analysis
BPM & Key Detection
Every track gets automatic tempo (BPM) and musical key detection on upload. Perfect for DJs mixing harmonically, producers matching samples, or musicians transposing covers.
- DJ-ready
DJ Mode & Rekordbox Export
Get beat grids, Camelot notation for harmonic mixing, and one-click Rekordbox XML export. Drop your stems straight into your DJ workflow — no extra setup required.
- Try before download
Waveform Stem Preview
Audition every stem in the browser before you download — mute, solo, and scrub with a synced waveform player. Hear the quality first, then export only what you need.
- < 60 seconds
Lightning Fast Processing
Get your separated stems in under 60 seconds. GPU-accelerated AI processes most songs faster than real-time — no waiting, no queues.
- Pay-as-you-go
Credits Never Expire — No Subscription
Start free, go Pro at $9/month, or buy a credit pack and pay only for what you process. Unlike subscription services that burn unused minutes every month, your credits stay yours forever.
- No card needed
10 Free Minutes — No Credit Card
Create an account and get 10 free minutes — enough to process 2–3 full songs in 6-stem quality. No credit card required, no trial countdown.
- 6+ formats
All Major Formats
Upload MP3, WAV, FLAC, M4A, OGG, or WEBM. Download each stem as a lossless WAV master, a compact MP3, or studio-grade FLAC — ready for your DAW, DJ software, or phone.
- 6
- Clean stems per upload
- < 60 sec
- For a 3-min song on GPU
- 20 min
- Longest track supported
- $0
- Cost of your first split
6-Stem Separation: Vocals, Drums, Bass, Guitar, Piano, and Other
AI Stem Splitter delivers six clean stems from a single track, while most free stem splitters cap out at four. You get isolated vocals, drums, bass, guitar, piano, and a residual "other" layer, giving you full stem-level control for remixes, mashups, karaoke tracks, and sample-ready loops.
- Vocals
- Drums
- Bass
- Guitar
- Piano
- Other
看得见的分离质量
我们的 6 轨输出把人声频率分离得干净；典型的 4 轨基线模型会留下鼓点/bass 的残留（频谱上可见的横向条纹）。
4 轨基线
htdemucs 6 轨（我们）
两条人声 stem 来自同一段 20 秒 Pop 素材。
htdemucs
Studio-Grade Separation Powered by htdemucs
Every split runs on htdemucs, the state-of-the-art hybrid transformer architecture developed by Meta AI and the winning model of the Sony Music Demixing Challenge. You get the same studio-grade separation quality that professional engineers rely on, with crisp vocals, tight low-end, and minimal bleed across instruments.
- Meta AIHybrid transformer architecture
- Sony MDX WinnerMusic Demixing Challenge champion
Remix, Sample, and Remove Vocals in Minutes
You finally have stem-level control over any song you love. Upload a track, let the AI isolate vocals, drums, bass, and guitar in minutes, then flip your idea into a remix, a mashup, or a clean karaoke bed. Your creativity leads, your favorite songs follow.
How to Split a Song with AI Stem Splitter
Turning any song into six clean, ready-to-mix stems takes only three quick steps.
- 1
Upload your audio
Drop in any MP3, WAV, or FLAC track you want to separate.
- 2
Let the AI work
The model splits it into six studio-quality stems in about 2–3 minutes.
- 3
Preview and download
Play each track in the multi-track player and download the stems you need.
为制作人、DJ、声乐爱好者和 KTV 粉丝而生
制作人
采样鼓点、分离循环，从 stem 重新搭建整首歌。
DJ
提取干净的 acapella，为现场混搭添砖加瓦。
声乐爱好者
用独立人声轨练习和声、学习旋律。
KTV 粉丝
去掉人声，跟唱任何你爱的歌。
Demo audio: CC BY 3.0 “Traveling Lights” by Zep Hurme · CC BY 4.0 “Dusty Highway” by Sascha Ende
价格
按实际处理时长付费
登录即送 10 分钟，之后按需购买分钟包，用多少扣多少。
50 分钟
≈ 16 首歌
$6.99
适合先处理几首完整歌曲
最受欢迎
150 分钟
≈ 50 首歌
$15
适合经常拆歌的用户
500 分钟
≈ 166 首歌
$39.99
适合高频、重度使用场景
Frequently Asked Questions
Everything you need to know before your first split.
Start Your First Free Split
Upload a song right now and hear your first 6 stems in minutes — no watermarks, no catches. Love what you get back? Pay-as-you-go credit packs take over when your free credits run out, with no subscription and credits that never expire.
- Free credits to start
- Pay-as-you-go packs
- Credits never expire