How I Built a Full Content Pipeline Using n8n + Claude in One Weekend

I kept seeing the same problem: creators spending 4–6 hours repurposing a single long video into short-form content. The research, the scripting, the caption writing, the scheduling — all manual, all repetitive, all automatable.

So I built a pipeline to handle it. This is the full breakdown — tools, workflow, prompts, and what I'd do differently.

What the Pipeline Does

Input: a YouTube URL.
Output: 5 short-form scripts, captions, hashtag sets, and posts queued for scheduling. No manual work after the URL is dropped in.

The full flow:

1. Fetch YouTube transcript via API
2. Send to Claude with a segmentation prompt — identify 5 "clip-worthy" moments
3. Generate a short-form script for each (hook, body, CTA)
4. Write captions and hashtags per platform (TikTok, Instagram, LinkedIn)
5. Store everything in Notion
6. Queue posts via Buffer or a direct API

Tools Used

• n8n — workflow orchestration (self-hosted, free) *affiliate
• Claude API — script generation and caption writing *affiliate
• YouTube Data API — transcript fetch (free tier)
• Notion API — content storage and review queue
• Buffer API — scheduling (optional, can swap for any scheduler)

The n8n Workflow

The workflow has 6 nodes. Here's what each one does:

Node 1 — Webhook trigger. Accepts a POST request with a YouTube URL. You can trigger this from Notion, a form, or manually.

Node 2 — YouTube transcript fetch. Hits the YouTube Data API to pull the full transcript. If captions aren't available, it falls back to Whisper via a secondary node.

Node 3 — Claude segmentation. Sends the transcript to Claude with this prompt:

"You are a short-form content strategist. Read this transcript and identify 5 segments that would work as standalone short-form videos (60 seconds or less). For each segment, return: the timestamp range, a one-line hook, and why this segment works as a standalone clip."

Node 4 — Script generation loop. Iterates over the 5 segments. For each, sends another Claude call with the segment text and generates a full short-form script: hook (first 3 seconds), body, and CTA.

Node 5 — Caption + hashtag writer. Takes each script and generates platform-specific captions. One prompt, three outputs (TikTok, Instagram, LinkedIn). Hashtags are generated separately and appended.

Node 6 — Notion write. Saves everything to a Notion database with status "Review". You review, approve, and the next automation picks it up for scheduling.

The Claude Prompts

The prompts are the most important part. Bad prompts = generic output. Here's the script generation prompt that works:

"Write a short-form video script for this segment. The script must: start with a pattern-interrupt hook in the first 3 seconds, deliver one clear insight or story in under 60 seconds, and end with a soft CTA (no hard selling). Tone: direct, conversational, no fluff. Format: [HOOK] / [BODY] / [CTA]"

Results From First Run

I ran this on a 45-minute finance creator interview. Output in under 4 minutes:

• 5 scripts, all usable with minor edits
• 15 captions (3 platforms × 5 scripts)
• 5 hashtag sets
• Everything in Notion, ready to review

Time saved vs. manual: ~5 hours per video.

What I'd Do Differently

The transcript quality matters a lot. Auto-generated YouTube transcripts are messy — lots of run-on sentences, missing punctuation. Adding a cleaning step (another Claude call to reformat the transcript before segmentation) improved output quality significantly.

Also: the scheduling node is the weakest link. Buffer's API has rate limits and the free plan is restrictive. If you're doing volume, build directly to the platform APIs.

Next Steps

I'm adding a video clipping node using FFmpeg to auto-cut the actual video segments based on the timestamps. When that's done, the pipeline will output both the script and the raw clip — ready to hand off for editing or drop directly into CapCut.

I'll document that build in the next post.

---

If you want this pipeline built for your workflow, reach out. The AI Automation Setup service starts at $2,000 and includes full documentation and a 30-minute handoff call.