Automating Short-Form Video Production with Claude Code

Duncan Rogoff | AI Automationgo watch the original →

A workflow using Claude Code to ingest long-form YouTube transcripts, generate three distinct short-form scripts based on proven hook patterns, create AI avatars via HeyGen, and automate multi-platform scheduling.

Automated Content Repurposing Workflow

The author uses Claude Code to transform long-form YouTube videos into short-form content by automating the extraction of key moments, script generation, and asset assembly. The system operates by ingesting a YouTube URL, parsing the transcript with timestamp alignment, and applying specific content strategies to generate three distinct angles: proof, contrarian, and transformation.

Prompting and Scripting Strategy

To ensure high-performing hooks, the system references a vault of proven patterns rather than generating new ones. The author mandates that the model swaps variables within these established structures. The script generation process follows a strict hierarchy: lead with the result, name the specific pain point removed, and maintain a natural, conversational cadence. The system is instructed to prioritize existing screen captures from the source video as the primary B-roll, only generating new AI images when necessary to fill conceptual gaps.

Avatar and Asset Integration

Digital clones are created using HeyGen by uploading a high-quality image of the creator, which is then styled via Google AI Studio prompts to match a professional studio aesthetic (shallow depth of field, warm softbox lighting). The system integrates with the HeyGen API to automate the lip-syncing of the generated scripts. Final output is managed through a third-party scheduling tool, which distributes the content across Instagram, TikTok, YouTube Shorts, and LinkedIn on a staggered schedule.

  • #ai
  • #automation
  • #content-creation

summary by google/gemini-3.1-flash-lite. probably wrong about something. check the source.