Content Repurposing: Turn One Podcast into 20 Posts
One podcast, 20 pieces of content. Here is how to build AI agents that auto-generate TikTok clips, tweets, LinkedIn posts from one video.
You recorded a podcast. It took two hours of prep, one hour of recording, another hour of editing. Then you posted it and watched it get 200 views.
Meanwhile, some 22-year-old with a ring light is pulling millions of views by reposting the same content in six formats across eight platforms.
The difference is not talent. It is that they understood: one piece of content is raw material, not a finished product.
The problem has always been the labor. Chopping a one-hour podcast into clips, writing tweet threads, reformatting for LinkedIn — that is a full-time job. Which is why most people do not do it.
But now you can build an AI agent that does the entire thing in minutes.
The Pipeline
Stage 1: Transcription
Everything starts with turning audio into text. OpenAI Whisper is free, runs locally, 95 percent accurate. AssemblyAI handles multiple speakers and gives word-level timestamps — critical for cutting video.
Stage 2: Hook Detection
Feed the transcript into an LLM and ask it to score segments by virality potential. Look for contrarian takes, specific numbers, emotional peaks, quotable one-liners.
The LLM outputs a JSON array of segments with timestamps, scores, and suggested captions. This alone is worth the entire build.
Stage 3: Multi-Format Generation
Same content, completely different wrappers. Twitter needs punchy one-liners. LinkedIn rewards vulnerability plus lesson structure. TikTok needs scroll-stopping first lines.
Use CrewAI or LangChain to orchestrate multiple specialized agents — one for Twitter, one for LinkedIn, one for clipping.
Stage 4: Video Clipping
Opus Clip or 2short.ai auto-detect best moments, add captions, reframe for vertical. Or build it yourself with FFmpeg and whisper timestamps.
Stage 5: Scheduling
Integrate with Buffer, Typefully, or direct API integrations. Space out clips over 10-14 days for daily posting.
The Stack
- Transcription: AssemblyAI
- Orchestration: CrewAI
- LLM: Claude or GPT-4
- Video clipping: Opus Clip
- Scheduling: Buffer API
Monthly cost: $50-100. Compare that to $1,500-3,000 for a VA.
Where This Makes Money
First — your own content. Consistent multi-platform content is the best acquisition channel. This removes the production bottleneck.
Second — as a service. Podcasters will pay $500-2,000/month for repurposing. Deliver with AI, do quality control.
Third — as a product. Wrap a UI around the pipeline and sell as SaaS.
Common Mistakes
Do not skip platform-specific formatting. Do not publish without reviewing — AI gets you 85 percent there. Do not ignore the hook — first 1-3 seconds determine everything.
Next Steps
Pick one piece of content you have already created. Run it through Whisper. Feed the transcript to Claude with hook detection. Generate three posts. Cut one clip and post it.
That is your proof of concept.
The era of create once, post once is over. Creators who build these systems will own the next five years of audience growth.