Drop a node. Split any audio.
Add stem splitting to any n8n workflow in under five minutes — no glue code, no servers, on n8n Cloud or self-host. The maintained community node ships async operations (Submit Split, Wait for Split, plus a dedicated AI Stem Splitter Trigger for webhook receivers) so long-running jobs never time out the execution.
Three workflows the node already supports end to end
Auto-stem Dropbox uploads
Watch a Dropbox folder for new tracks. Each upload runs Submit Split, waits on completion, and writes the four-stem set back to a per-track subfolder in Drive or Dropbox.
Daily karaoke playlist (YouTube cron)
A Schedule (cron) trigger pulls the day's YouTube uploads from a channel, sends each track through Submit Split with the vocals stem suppressed, and pushes the karaoke version into a public playlist or CDN bucket.
Slack stem bot
A Slack slash command posts an audio link. The AI Stem Splitter Trigger node receives the completion webhook and replies in-thread with download links for each stem.
Install on n8n Cloud or self-host
Run npm install n8n-nodes-aistemsplitter from your custom-nodes directory, then restart n8n. Works on Docker, npm, and desktop self-host installs. MIT licensed.
# 1. Install the package
npm install n8n-nodes-aistemsplitter
# 2. Restart your n8n instance
n8n start
# 3. The "AI Stem Splitter" node appears in your nodes panel.Open Settings → Community Nodes, click Install, paste n8n-nodes-aistemsplitter, and confirm. Available on Cloud tenants where community nodes are enabled.
# In n8n Cloud:
# 1. Settings → Community Nodes → Install
# 2. npm Package Name: n8n-nodes-aistemsplitter
# 3. Tick "I understand the risks", click Install
# 4. Refresh the workflow editor — node now in the panelYour first workflow in 5 steps
Add a trigger
Pick the trigger that matches your workflow: Webhook for on-demand splits, Schedule for cron jobs, Google Drive or Dropbox for upload-driven runs, or Slack for chat-driven splits. The trigger feeds an audio URL into step 3.
Create the credential
Open Credentials → New, search for AI Stem Splitter API, paste your ast_live_ key, and save. The node verifies the key against the API and shows ✓ Connected. Same credential is reused by every AI Stem Splitter node in this instance.
Add the AI Stem Splitter node
Drop the AI Stem Splitter node onto the canvas, choose operation Submit Split, and fill the three fields: audioUrl (from the trigger), model (default htdemucs_ft, four stems — switch to a six-stem model when you need guitar or piano), and stems (chip-pick the stems to return).
Wait for completion
Stem splitting takes 30 seconds to a few minutes per file. Pick one of two patterns: the Wait for Split operation pauses this execution until the job finishes, or the dedicated AI Stem Splitter Trigger node receives the completion webhook in a separate workflow — better for high-volume queues that should not block the main run.
Use the stems
The output is an array of stem URLs (vocals, drums, bass, other; plus guitar and piano when requested). Wire the URLs into your downstream nodes — Whisper for vocal-only transcription, Drive or S3 for storage, Slack for a chat reply, a Notion row for a stem library entry.
Templates and further reading
FAQ
Does the n8n node work on n8n Cloud or only self-host?
Both. On self-host, run `npm install n8n-nodes-aistemsplitter` from your custom-nodes directory and restart n8n. On n8n Cloud, open Settings → Community Nodes, paste n8n-nodes-aistemsplitter, and confirm — community-node install is supported on all current Cloud plans (legacy tenants may need to enable community nodes in workspace settings first).
How does the workflow handle long-running jobs without timing out?
Two patterns. The Wait for Split operation pauses the current workflow until the job finishes (good for low-volume queues). The dedicated AI Stem Splitter Trigger node receives the completion webhook in a separate workflow (better for high volume — main workflow returns immediately, downstream actions fire when stems are ready). Stem separation typically takes 30 seconds to a few minutes per file.
Does the trigger node work behind a firewall or behind a VPN?
The Trigger node needs an inbound webhook URL the API can POST to. On n8n Cloud, the trigger URL is automatically public. On self-host behind a firewall, use n8n's built-in tunnel (n8n start --tunnel) for testing, or expose the webhook via a reverse proxy / Cloudflare Tunnel for production. If inbound webhooks aren't possible at all, fall back to the Wait for Split operation (polling, no inbound).
Can I export and share a workflow that uses this node?
Yes. Use n8n's standard Workflow → Download to export the workflow as JSON. Credentials are referenced by ID, not embedded — recipients re-create the AI Stem Splitter API credential with their own ast_live_ key when they import. The same workflow JSON works across self-host and Cloud as long as the recipient also installs n8n-nodes-aistemsplitter.
Will the node break when n8n core upgrades?
Community nodes pin against n8n's stable interfaces. We test against each new minor n8n release and publish a matching version of n8n-nodes-aistemsplitter to npm within a few days when needed. Pin a specific node version in your custom-nodes package.json if you want zero surprise; the GitHub repo's release notes call out any breaking change explicitly.
How is per-job cost calculated when the workflow runs on every upload?
Per minute of input audio against credit packs that never expire — $0.08–$0.14 per minute depending on pack size. A 4-minute song is ~$0.32–$0.56; a 60-minute podcast is ~$4.80–$8.40. The first 3 minutes are free with no card required. There's no per-call surcharge, no concurrency tier gate, and no monthly subscription — only audio you actually process.
Ship the stem step before EOD.
Free 3 minutes to start (no card). Credit packs that never expire — $0.08–$0.14 per minute depending on volume. MIT licensed node, on n8n Cloud or self-host.