AI Stem Splitter API
aistemsplitter.org
- Free plan
- Windows
- macOS
- Web
- iPhone and iPad
- Android
- API
What AI Stem Splitter API is
AI Stem Splitter API separates audio into vocals, drums, bass, guitar, piano, and other stems. It can remove vocals for karaoke instrumentals or isolate acapellas for remixing. Users can upload MP3, WAV, FLAC, M4A, AAC, OGG, or WEBM files up to 50 MB, or submit YouTube and SoundCloud links. Tracks can be up to 20 minutes. Stems can be previewed in a browser waveform player and downloaded as WAV, MP3, or FLAC. The service says it uses htdemucs, with htdemucs_6s for guitar and piano. DJ mode provides beat grids, Camelot notation, and Rekordbox XML export. The REST API accepts URLs or uploads, supports job-status polling, and offers signed webhooks. SDKs are listed for Node.js, Python, Java, Go, PHP, Swift, and Lua, with n8n, GitHub Actions, Docker/CLI, and Pipedream options. A free trial provides three processing minutes without a credit card. Paid packs start at $6.99, paid once for 50 minutes, and purchased minutes never expire.
Who it suits
AI Stem Splitter API suits producers, DJs, vocalists, karaoke users, and developers integrating stem separation through an API.
Good to know, and what to watch
- Separates five or more stem types
- YouTube and SoundCloud link processing
- WAV, MP3, and FLAC exports
- REST API with signed webhooks
- Three-minute free trial
- Tracks limited to 20 minutes
- Uploads limited to 50 MB
- Paid packs start at $6.99, paid once
Plans
Every plan and priceFree trial
Free trial — 3 mins
50 mins
50 minutes
The verdict
AI Stem Splitter API combines browser-based stem work with developer integrations and DJ-oriented exports. Track, upload, and minute-pack limits should be checked against the intended workflow.
What the maker publishes
- API
- The REST API supports direct URLs or uploads, job-status polling, and signed webhooks. source
- API authentication and credits
- API requests use bearer API keys, and one credit represents one second of source audio. source
- API errors
- Documented errors include unauthorized, insufficient credits, validation, rate limiting and server errors with HTTP codes 401, 402, 422, 429 and 500. source
- API input and status flow
- The API supports direct URLs or presigned uploads, then exposes queued, processing, succeeded and failed job states. source
- Audio length limit
- The homepage lists 20 minutes as the longest supported track. source
- Audio model
- The site says its separation uses htdemucs, with guitar and piano provided by the htdemucs_6s model. source
- Billing unit
- Billing deducts the exact duration of submitted audio, measured down to the second. source
- BPM and key detection
- Uploads receive automatic tempo and musical-key analysis. source
- Credit expiry
- Both free and purchased minutes remain available until consumed and do not expire. source
- Developer integrations
- The SDK page lists first-party SDKs for Node.js, Python, Java, Go, PHP, Swift, and Lua, plus n8n, GitHub Actions, Docker/CLI, and Pipedream options. source
- DJ export
- DJ mode provides beat grids, Camelot notation and Rekordbox XML export. source
- DJ features
- DJ mode offers beat grids, Camelot notation, and Rekordbox XML export. source
- Download formats
- Separated stems can be downloaded as WAV, MP3 or FLAC files. source
- Exports and preview
- Users can preview stems in a browser waveform player and download WAV, MP3, or FLAC files. source
- Failed-job refunds
- The service pre-charges a submission and automatically refunds the same amount when processing fails. source
- Formats and size
- The upload form lists MP3, WAV, FLAC, M4A, AAC, OGG, and WEBM, with a 50 MB maximum. source
- Input formats and size
- The homepage accepts MP3, WAV, FLAC, M4A, AAC, OGG and WEBM files up to 50 MB. source
- Intended users
- The homepage presents the tool for producers, DJs, vocalists, and karaoke fans. source
- Limit: Cyrillic filenames
- A forum participant said files whose names use Cyrillic characters failed to upload. source
- Limit: pricing criticism
- One forum reply accused the service of charging for software the commenter believed was freely available and called that deceptive. source
- Limit: regional access
- A forum participant reported that the site was inaccessible from their region. source
- Limit: reported glitches
- The same discussion includes a user who said uploads sometimes produced errors and described the site as buggy. source
- Link processing
- Users can submit YouTube or SoundCloud links for direct processing into stems. source
- n8n capabilities
- The verified n8n node offers Create Split, Get Credits and Get Split actions. source
Checked against the maker's pages on 15 Oct 2026.