The n8n content machine: your channel clips itself, you approve from your phone
Your first video is free up to 60 minutes, then you get 20 free minutes a month, with no card requested.
This workflow runs a YouTube channel on its own without handing over the publish button. Once a day it reads your channel feed, clips one video into vertical shorts, and sends each finished clip to your Telegram with Publish and Skip buttons. What you approve is scheduled one post per day to your connected accounts, and every Sunday it reports back what the published clips actually did. Two credentials, no polling, no TikTok or Instagram OAuth app.
Building the workflow
Every step below is a stock n8n node, so the machine can be rebuilt in the editor: Schedule, HTTP Request, Webhook, Telegram and HTTP Request again for publishing.
What the workflow actually does
Four stages, all on one canvas:
- Watch the channel. A daily schedule trigger reads
youtube.com/feeds/videos.xml?channel_id=..., which needs no YouTube API key, and picks one video: your newest upload, or, if there is nothing new, the next unprocessed video from your back catalogue. One video a day keeps the minute burn predictable, and a 402 (out of minutes) pauses the machine with a Telegram notice rather than failing silently. - Clip, then get called back. One HTTP request to
/api/processcarrying awebhook_url. When the job ends, Clipping AI POSTs once with the finished clips and durable download links. There is no Wait node anywhere in the workflow. - Approve from your phone. Each 9:16 clip arrives in Telegram as a video message with Publish and Skip buttons. A human approves every post, which is also what separates this from the fully automated pipelines that YouTube's inauthentic-content policy targets.
- Drip-publish and measure. Approved clips take the next free daily slot and post to the accounts you connected in Clipping AI. Every Sunday the workflow reads back the analytics of what it published and sends you impressions, the per-platform split and your best post.
Why posting needs no TikTok or Instagram credentials
The usual wall in an n8n video workflow is the publishing half: TikTok
requires an audited app to post publicly, and the Instagram Graph API refuses
anything that is not a public static URL, which is why Google Drive links fail
there. This workflow sidesteps both by posting through POST /api/social/post to the networks you connected once in your Clipping AI account. The workflow itself holds no social credentials, and
the clip file is already on durable storage, so the public-URL requirement is met before Instagram ever sees it.
Scheduling that will not double-book
Approving two clips seconds apart used to be enough to publish them in the same minute: each execution read the same in-memory counter before either wrote back. The workflow now asks the server for the queue it actually holds
(GET /api/social/scheduled) and picks the first free slot from it,
which is shared state and cannot race with itself that way. That endpoint, plus DELETE /api/social/scheduled/{job_id}, is how you inspect or cancel a post before it goes out.
What it costs to run
Aim it at Clipping AI Cloud and it runs on the first video clipped whole up to 60 minutes, then 20 minutes a month with a watermark and no card, then on a plan from $12/month; billed yearly the entry plan is $10/month.
API calls draw from the same minute balance as the dashboard: no per-call price, no automation surcharge. Telegram bots are free, and n8n runs wherever you already run it.
Known limits, before you hit them the hard way
- Telegram previews a video by URL up to about 20 MB; larger clips arrive as a link message carrying the same approval buttons.
- A YouTube channel RSS feed exposes only the latest 15 videos, so the back-catalogue drip reaches back that far and no further.
- The Telegram Trigger node needs your n8n to have a public https URL, because Telegram registers a webhook against it. A laptop-local n8n can run every other stage, but the approval buttons need a deployed instance or a tunnel.
- Workflow static data, where the machine remembers which videos it already processed, only persists on production runs. Test runs from the editor do not advance it.
Prefer an agent to a workflow?
The same account exposes an MCP server, so Claude, ChatGPT or an n8n AI Agent node can drive the pipeline as tools instead of fixed steps. That surface is documented on the MCP server and API page, and the raw REST loop on the automation page.
Questions people ask
- Is there a free n8n template for turning long videos into shorts?
- Yes. This page walks through an n8n workflow that clips a YouTube channel automatically and posts the approved clips to TikTok, Instagram and YouTube, built entirely from stock n8n nodes.
- How do I post a video to TikTok or Instagram from n8n?
- Neither platform has a native n8n node, and both have hard requirements: TikTok needs an audited app for public posts and Instagram needs a public static URL for the media. Posting through the Clipping AI API avoids both, because the accounts are connected once in your Clipping AI account and the clip already lives on durable public storage.
- Does the workflow poll for the clipping job to finish?
- No. Clipping a real video takes minutes, so the workflow passes a webhook_url with the job and Clipping AI calls it exactly once when the job reaches a terminal state. Failed jobs fire the same webhook with an error field, so the flow never hangs.
- Can the clips publish without my approval?
- They can, by connecting the posting step directly to the webhook branch, but the template ships with the approval gate on purpose: unreviewed automated publishing is what YouTube's inauthentic-content policy targets, and one bad clip lands on your own audience.
Where these numbers come from
- TikTok and Instagram publishing constraints checked against their developer documentation, August 2026.