The barrier between raw footage and published content is crumbling. NJ, a developer and founder of SimbaStack, recently demonstrated that Claude can autonomously edit YouTube videos from start to finish using DaVinci Resolve’s new MCP (Model Context Protocol) connection. Introduced in Resolve 21.1, this integration allows AI agents to control the editing software directly, marking a significant shift from simple transcription tools to full creative execution.

The Workflow: From Raw Files to Render

NJ’s process is deceptively simple but technically robust. After recording raw clips, he drops them into a folder and provides Claude with a voice-dictated 'thought dump' via the Handy app. The AI then handles the heavy lifting: transcribing audio with Whisper, cutting out filler words like 'um' and pauses, and selecting the best takes. For complex decisions, such as choosing between multiple takes, Claude generates A/B timelines and asks for simple alphanumeric inputs (e.g., '1A 2C') from the user. The system also generates motion graphics, music beds, and thumbnail options using Python scripts and ffmpeg outside the Resolve environment before importing the final assets for rendering.

Performance Metrics and Cost Efficiency

The efficiency gains are stark. NJ’s first video required a full day and 11 timeline versions to complete. By his third video, the entire process took just 57 minutes, using only 2–3% of his weekly limit on the Claude Max (20x) plan. The final cut of the third video, a 10:35-minute piece derived from 28:42 of raw footage, required only five timeline versions and minimal human notes. This suggests that as the agent 'learns' the user’s style through reusable scripts, the time cost per video drops precipitously, potentially undercutting the $300 cost of hiring a freelance editor for a single project.

Technical Hurdles and AI Autonomy

Despite the success, the integration is not without friction. NJ encountered issues with audio clipping and frame rate mismatches (24 fps default vs. 29.97 fps footage), which required manual intervention in Project Settings because the scripting API cannot adjust certain core parameters. Additionally, early versions of the workflow risked overwriting manual trims when rebuilding timelines, a bug since patched by having the agent duplicate versions rather than rebuild from scratch. Interestingly, Claude also demonstrated fact-checking capabilities, flagging NJ’s incorrect description of baboons as herbivores and correcting him to omnivores, showing that the agent is parsing semantic content, not just following editing cues.

Key Takeaways

  • DaVinci Resolve 21.1's MCP integration enables Claude to perform end-to-end video editing, reducing production time from days to under an hour.
  • The workflow requires a one-time manual setup for frame rates and uses a hybrid approach where Resolve handles cuts while external scripts manage graphics and audio.
  • AI autonomy extends to semantic understanding, including fact-checking spoken content and adapting to user feedback via reusable scripts.
  • For non-professional creators, this pipeline offers a cost-effective alternative to hiring editors, utilizing less than 3% of a Claude Max weekly limit per video.

The Bottom Line

While AI won’t replace high-end creative editing anytime soon, this workflow proves that for 'normie' content creators, the technical drudgery of editing is now automatable. The combination of Resolve’s scripting API and LLM reasoning creates a powerful, low-cost pipeline that turns video production from a time sink into a manageable task.