Communitygithub.com

liuhaodong/ai-video-editor

Modular multi-agent AI photo and video editing skill for family travel, vlogs, and unscripted highlights. Orchestrates specialized sub-agents (Client Intake, Culler & Sync, Audio/Beat Vibe Designer, Story Producer, Dual Dialogue & Montage Editors, Pluggable VFX/Color Agents, Showrunner Pacing Critic, and Executive Movie Production Director) through deterministic JSON contracts (creative_brief.json, media_index.json, beat_grid.json, paper_edit.json, pacing_audit.json, director_review.json, timeline.yaml) and renders 4K 10-bit HLG HDR deliverables + FCPXML/EDL/OTIO.

What is ai-video-editor?

ai-video-editor is a Antigravity agent skill that modular multi-agent AI photo and video editing skill for family travel, vlogs, and unscripted highlights. Orchestrates specialized sub-agents (Client Intake, Culler & Sync, Audio/Beat Vibe Designer, Story Producer, Dual Dialogue & Montage Editors, Pluggable VFX/Color Agents, Showrunner Pacing Critic, and Executive Movie Production Director) through deterministic JSON contracts (creative_brief.json, media_index.json, beat_grid.json, paper_edit.json, pacing_audit.json, director_review.json, timeline.yaml) and renders 4K 10-bit HLG HDR deliverables + FCPXML/EDL/OTIO.

Works with~Claude Code~Codex CLI~Cursor✓Antigravity
npx skills add https://github.com/liuhaodong/video_editing_ai/tree/HEAD/.agent/skills/ai-video-editor

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Modular Multi-Agent Video & Photo Editor Skill

Use this skill whenever the user wants to analyze a folder of videos and photos, interview/clarify story preferences, run the multi-agent vlog & family travel editing pipeline, add a custom editing sub-agent role (such as VFX or Color), or render a 4K 10-bit HLG HDR video.

1. Sub-Agent Role Reference Files (Progressive Disclosure)

Each agent role has its own reference specification in references/ (used by both Antigravity's define_subagent / invoke_subagent tools and the Python BaseEditingAgent classes):


2. Running the Multi-Agent Pipeline

Option A: Native Antigravity Subagents + CLI Orchestrator

  1. Interview the Client (Stage 0): Use ask_question (guided by references/00_client_intake_agent.md) to clarify narrative style, must-have moments, target duration, and photo framing, or run:
    python -m video_editing_ai.cli agents-run "<media_dir>" -o "<project_dir>" --prompt "<user_story_idea>" --bgm "<optional_bgm.mp3>"
    
  2. Spawn Specialized Subagents (define_subagent / invoke_subagent): You can spawn any of the roles in isolated context windows using define_subagent with the system prompt from its corresponding references/*.md file so each role inspects and enriches its JSON contract on disk (creative_brief.json → media_index.json & beat_grid.json → paper_edit.json → timeline.yaml → pacing_audit.json).
  3. Render 4K 10-bit HLG HDR Video & NLE Exports:
    python -m video_editing_ai.cli render "<project_dir>/timeline.yaml" -o "<project_dir>/final_story.mp4"
    

Option B: List or Toggle Pluggable Sub-Agents

# List all registered core and optional plugin agents
python -m video_editing_ai.cli agents-list

# Run the pipeline with the optional Stage-40 VFX Designer plugin enabled
python -m video_editing_ai.cli agents-run "<media_dir>" -o "<project_dir>" --prompt "<idea>" --enable-agent vfx_designer

Option C: Local Web Studio

python -m video_editing_ai.cli serve --port 8000

Related Skills