Podcast to Shorts: Automated Clips with Whisper + CapCut

Turn every podcast episode into 5–10 Shorts automatically. Whisper for transcripts, GPT for highlights, CapCut templates for on-brand clips.

Pipeline Overview

1. Transcribe: Whisper large-v3 (local or API) to get timestamped text.
2. Highlight mining: GPT extracts Q&A, emotional peaks, quotable lines + timecodes.
3. Clip assembly: CapCut template with waveform, big captions, brand colors.
4. B-roll: Runway/Pexels CC0 for visual variety; keep 9:16.
5. Publish: YT Shorts + TikTok + Reels with platform-specific titles.

Prompts & Settings

• Highlight prompt: “Find 5–10 moments with conflict, humor, ‘how to’, or strong opinion. Return start/end timestamps and a 10-word hook.”
• Caption style: sentence case, high contrast, 2–3 lines max; place lower-third.
• Audio: loudness target -14 LUFS; compress lightly, remove breaths/filler with Descript.

Brand Consistency

• Lock template colors, logo, and lower-third position; avoid per-clip design drift.
• Add a 3-second outro CTA to full episode with a shortened link/QR.

Publishing Playbook

• Titles: “Question + surprising word” for Shorts; TikTok uses emoji/emotion; IG uses action verbs.
• Descriptions: first line links full episode with UTM; add 3–5 niche tags.
• Spacing: stagger uploads over 24h; avoid dumping all clips at once.

Quality & Compliance

• De-noise before transcription; check for copyrighted music in the original.
• Respect guest likeness/contract if faces appear—blur or replace with graphics when needed.
• Check captions sync on 2–3 clips before batching.

Metrics to Watch

• 30s retention and replays; hook CTR by title variant.
• Click-through to full episode (description link); saves/shares per platform.
• Time-to-publish: target <30 minutes from audio to finished Short once templates are set.