Automating DaVinci Resolve Workflows with GPT-6 Astra

Lukas Margeriego watch the original →

The author demonstrates ten AI-driven workflows for DaVinci Resolve, using GPT-6 Astra and companion tools like HeyClicky to automate editing, color grading, motion graphics, and B-roll generation.

Automated Project Setup and Editing

The author uses GPT-6 Astra via the Codex integration to automate the initial assembly of raw footage. By attaching a video file to the prompt, the model performs a rough cut by removing silent segments and repeated phrases. To refine the timeline for social media formats, the author uses HeyClicky to generate specific prompts that force the model to reformat the timeline to vertical, center the subject, and close remaining gaps between clips.

Visual Enhancements and Motion Graphics

To improve visual quality and add motion elements, the author leverages specialized AI plugins and reference-based prompting:

  • Color Grading: The author uses the Higgsfield plugin within the Astra environment, providing a reference tweet to guide the model in applying a specific color grade, which the author notes results in sharper shadows and improved skin tones.
  • Transitions: By referencing URLs from Remotion or Hyperframes documentation, the author prompts the AI to insert specific transition effects at designated timestamps.
  • Layout Sketching: Using MagicPath, the author sketches custom layouts for picture-in-picture and green-screen effects, then prompts the AI to apply these layouts to specific clips.
  • UI Animation: The author sketches a UI overlay in MagicPath and prompts Diffusion Studio to render a 4-second animation of a cursor clicking a submit button, which is then overlaid onto the timeline.
  • Remixing Assets: The author takes existing product launch videos and prompts the AI to recreate the motion style while swapping text and colors to match a personal brand.
  • B-roll Generation: The author uses Seedance 2.5 to generate context-aware B-roll overlays that replace talking-head segments based on the audio content of the clip.

Context

The author explores how multimodal AI models can act as a direct interface for professional video editing software. By combining visual sketching tools with LLM-based command execution, the author demonstrates a workflow that moves beyond simple text-to-video generation, focusing instead on manipulating existing project timelines and assets within DaVinci Resolve.

  • #tutorial
  • #ai
  • #dev-tooling

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