Automatically detect Sex & Nudity content in movies using GPT-4o vision analysis. Uses IMDB-style categories to identify content for filtering.

Sex (any sexual nature):

• Kissing (passionate/romantic kissing, making out)
• Groping/touching (intimate touching, caressing)
• Sexual activity (intercourse, implied or explicit)
• Suggestive behavior (seductive poses, sensual dancing)

Nudity:

• Revealing clothing (low-cut tops, swimwear, lingerie)
• Partial nudity (exposed back, cleavage, shirtless in sexual context)
• Full nudity (breasts, buttocks, genitals)

1. Scene Detection - Uses PySceneDetect to find natural scene boundaries
2. Frame Extraction - Extracts 8 evenly-spaced frames from each scene
3. Vision Analysis - Sends frames to GPT-4o for NSFW content detection
4. Segment Output - Generates JSON with timestamps of flagged scenes

ummah/
├── process_movie.py      # Main CLI entry point
├── src/
│   ├── __init__.py
│   ├── config.py         # Configuration and .env loading
│   ├── utils.py          # Time formatting helpers
│   ├── scene_detector.py # PySceneDetect wrapper
│   ├── frame_extractor.py# Video frame extraction
│   ├── vision_analyzer.py# GPT-4o vision API
│   └── aggregator.py     # Segment merging
├── requirements.txt
├── env_example.txt       # Example .env file
└── README.md

cd ummah

# Install dependencies
pip install -r requirements.txt

Create a .env file with your OpenAI API key:

# Copy the example file
cp env_example.txt .env

# Edit .env and add your key
OPENAI_API_KEY=your-api-key-here

Or set it as an environment variable:

export OPENAI_API_KEY="your-api-key-here"

python process_movie.py movie.mp4

Creates segments.json with detected NSFW segments.

python process_movie.py movie.mp4 -o my_segments.json

Lower threshold = more sensitive (catches more, may have false positives):

python process_movie.py movie.mp4 --threshold 0.4

Higher threshold = less sensitive (fewer false positives, may miss some):

python process_movie.py movie.mp4 --threshold 0.8

python process_movie.py movie.mp4 \
    -o segments.json \
    --threshold 0.6 \
    --frames-per-scene 8 \
    --min-scene-length 2.0 \
    --max-workers 4 \
    --frame-width 512

[
  {
    "start": "00:12:45",
    "end": "00:13:22",
    "severity": "moderate",
    "reason": "passionate kissing, revealing clothing",
    "confidence": 0.85
  }
]

Approximate costs per movie (GPT-4o):

• Short film (30 min): ~$0.15-0.25
• Standard movie (2 hr): ~$0.35-0.60
• Long movie (3 hr): ~$0.60-1.00

Reduce false positives: Increase --threshold to 0.7-0.8

Improve accuracy: Use --frames-per-scene 12 and --threshold 0.5

Speed up processing: Use --max-workers 8 and --frame-width 384

• Python 3.10+
• OpenAI API key with GPT-4o access
• FFmpeg (for video processing via OpenCV)