Skip to content

broomva/content-engine

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Content Engine

Full-stack AI content studio — compile visual identity once, generate premium content at scale, distribute everywhere.

What It Does

Content Engine treats visual identity (brand DNA, character sheets, style guides) as compiled knowledge — analyzed once from raw assets via Gemini multimodal, persisted as structured Markdown with tool-specific prompt fragments, and referenced by every generation session.

COMPILE → GENERATE → POST-PRODUCE → DISTRIBUTE → MEASURE → REFINE

Architecture

Four skills, each handling one layer:

Skill Purpose
content-engine-dna Visual DNA Compiler — raw assets → compiled identity files
content-engine-cinema Cinematic Generation — start-frame doctrine, camera vocabulary, 25+ styles
content-engine-autopilot Browser Orchestration — Playwright-driven batch generation
content-engine-loop Content Loop — compounds existing distribution skills

Quick Start

1. Prerequisites

pip install google-genai    # Gemini multimodal analysis
brew install ffmpeg          # Video keyframe extraction
export GEMINI_API_KEY="..."  # Required
export FAL_KEY="..."         # For Nano Banana 2, Kling via fal.ai

2. Add Raw Assets

knowledge/raw/
├── brand-assets/{brand-name}/     ← Campaign photos, logos
├── character-refs/{char-name}/    ← Face photos, pose refs
└── style-inspiration/{style}/     ← Mood boards, references

3. Compile

python3 scripts/compile-dna.py              # Incremental
python3 scripts/compile-dna.py --force      # Full recompilation
python3 scripts/compile-dna.py --dry-run    # Preview
python3 scripts/compile-dna.py lint         # Health check

4. Generate (via Claude Code)

/content-engine generate --brand acme --character luna --scenes 5 --format reels
/content-engine campaign "Mediterranean lifestyle, 10 summer scenes, golden hour"

Knowledge Architecture

Inspired by Karpathy's LLM Wiki and MemPalace:

knowledge/
├── raw/        # Immutable source material (never modified by LLM)
├── compiled/   # LLM-compiled identity files (the "executable")
└── schema.md   # Compilation rules

raw/ is source code. compiled/ is executable. The LLM is the compiler.

Tool Priority Matrix

Task Best Tool Fallback
Cinematic start frame Soul Cinema (Higgsfield) Nano Banana Pro
Character consistency Nano Banana Pro SD + LoRA
Scene variation Weavy Nano Banana + prompt
Video from keyframe Veo 3.1 / Seedance 2.0 Kling
Motion transfer Kling Wan
Upscaling Topaz Gigapixel Real-ESRGAN

Extension Points

Extensions hook into the pipeline at defined stages: pre-generation, post-generation, post-production, distribution. See extensions/README.md.

Planned: OpenCaptions (intent-driven captions), ComfyUI MCP, LoRA Training.

Research Sources

Built from analysis of industry creators and knowledge management frameworks:

  • viznfr — Claude Code + Playwright autopilot, character consistency
  • ohneis652 — ComfyUI node pipelines, LoRA style-locking, Soul Cinema
  • Vidis AI — Structured curriculum, ReelEngine/PromptEngine
  • Karpathy LLM Wiki — Compile-then-query, active linting
  • MemPalace — Spatial hierarchy, AAAK compression

License

MIT — see LICENSE.

About

Full-stack AI content studio: visual DNA compiler, cinematic generation, browser autopilot, content loop

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors