Overview
A single podcast episode should power a week of content. This n8n workflow watches Dropbox for new audio, transcribes with Whisper, extracts the best 60-90 second moments for social clips, writes a long-form blog post from the transcript, and stages everything as a Notion page ready for review.
How it works
- Dropbox trigger on new MP3.
- OpenAI Whisper node for transcription with timestamps.
- Claude extracts 3 highlight clips (best hooks, self-contained, 60-90s).
- Second Claude call writes a 1200-word blog post from the full transcript.
- Notion page created with sections: transcript, blog draft, clip timestamps + hooks, cover-image prompt.
Benefits
- Every episode yields 4-5 deliverables instead of 1.
- Clip timestamps mean your editor spends time cutting, not scrubbing.
- Blog draft becomes an SEO play from an audio asset.
Use cases
- Solo podcasters without a producer.
- Agencies packaging content for client shows.
- Internal 'lunch and learn' recordings turned into knowledge base pages.
Step-by-step guide
Step 1: Watch Dropbox
Trigger on new file in /podcast/episodes. Filter to .mp3/.wav/.m4a.
Step 2: Chunk long files for Whisper
Whisper's file limit is 25MB. Split with ffmpeg (n8n Execute Command node) if larger.
Step 3: Transcribe with timestamps
response_format=verbose_json for per-word timestamps. Reassemble into paragraphs for the blog step.
Step 4: Clip extraction prompt
System: 'Return 3 clips as JSON: start, end, hook, why_it_works.'
Step 5: Blog post from transcript
Ask for a 1200-word post with an H1, 5 sections, and 3 tweetable quotes at the top.
Step 6: Notion assembly
Single page with structured sections; owner assigned; status=Draft. Editor takes it from there.
Example
Clip output: {start:'00:14:20', end:'00:15:42', hook:'The real bottleneck isn't compute, it's judgement', why:'Reframes a common assumption in 8 seconds.'}