How to Automate Podcast Clips to Social Media

A podcast episode is 60 minutes of quotable material that most shows promote with… one link post. The clips are sitting in the audio — extraction is the automatable part.

The Pipeline

Step 1 — Upload the Episode

Presigned upload for local files or upload_from_url if the episode is already hosted — either returns a permanent media-storage fileUrl.

Step 2 — Transcribe (Built In — No External Whisper)

{ "url": "https://mediastorage.dohoo.ai/file/dohoo-video-storage/ep42.mp4" }
// → { "text": "…full transcript…", "language": "en", "duration": 3720 }

Language auto-detects (or pin with ISO codes: "en", "uk", "es"). One call, no separate transcription vendor, no file shuttling. This is the step where DOHOO replaces an entire service in the typical podcast stack.

Step 3 — Mine the Transcript for Clip Moments

Feed the transcript to your LLM with a selection prompt:

"From this transcript, find 8 self-contained moments of 30–60 seconds that contain a specific number, a contrarian claim, or a story with a payoff. Return start/end timestamps and the verbatim hook line."

Concrete beats generic: "we cut churn 40% with one email" clips; "consistency matters" doesn't.

Step 4 — Cut and Publish

ffmpeg -i ep42.mp4 -ss 00:14:32 -to 00:15:18 \
  -vf "crop=ih*9/16:ih,scale=1080:1920" clip03.mp4

Upload each clip, then schedule the multi-format rollout — all from one episode:

Spread clips across the week following release — the episode promotes itself daily instead of once.

Step 5 — Verify

Weekly audit of scheduled posts (status: "all"): every clip shows published or failed with an errorMessage. A 20-post rollout with one silent failure is a hole in the campaign; here failed scheduled records can be detected and reviewed through status checks.

Output Per Episode

One 60-minute episode → 6 clips × 4 video surfaces + 3 text posts + 1 pin ≈ 28 posts.

FAQ