ai-cartoon-cam

A local real-time camera stylization project for macOS, designed for OBS and livestream workflows.

Quick Start

1) Clone and enter project

git clone git@github.com:cowbook/ai-cam.git
cd ai-cam

2) Create environment and install

python -m venv .venv
source .venv/bin/activate
pip install -e '.[virtualcam,dev]'

3) Run real-time preview

ai-cartoon-cam --style cartoon

Try another style:

ai-cartoon-cam --style beauty

4) Run with config

ai-cartoon-cam --config examples/config.sample.yaml

5) Output to virtual camera for OBS

ai-cartoon-cam --style beauty --virtual-cam

Then in OBS, select the pyvirtualcam device as your camera source.

Vision

Build an open-source, continuously improving project for:

• Real-time cartoon/sketch/anime style effects.
• Smooth macOS local inference and camera processing.
• OBS virtual camera output for livestreaming platforms.
• A plugin-friendly architecture for future AI models.

Current features (v0.1.0)

• Real-time webcam capture via OpenCV.
• Built-in styles: cartoon, sketch, anime, beauty, raw.
• Optional virtual camera output via pyvirtualcam.
• YAML config support.
• CLI entrypoint for fast experiments.

Project structure

ai-cartoon-cam/
  src/aicartooncam/
    cli.py
    config.py
    filters.py
    pipeline.py
  tests/
  examples/
  docs/

Install

cd /Volumes/T4/2026/ai-cartoon-cam
python -m venv .venv
source .venv/bin/activate
pip install -e '.[virtualcam,dev]'

Run

Basic preview

ai-cartoon-cam --style cartoon

ai-cartoon-cam --style beauty

Use config file

ai-cartoon-cam --config examples/config.sample.yaml

Save and load presets

Save a merged preset (defaults + config + CLI overrides):

ai-cartoon-cam --config examples/config.sample.yaml --style anime --save-preset stream-anime

Load a preset by name on next run:

ai-cartoon-cam --preset stream-anime

By default presets are stored in ~/.ai-cartoon-cam/presets. You can override with --preset-dir.

Output to OBS virtual camera

ai-cartoon-cam --style beauty --virtual-cam

Then in OBS, choose the virtual camera device produced by pyvirtualcam.

AI Session Continuity

To continue work across new chats/sessions, start with:

• aiwork/memory.md
• aiwork/continue-prompt.md

Update aiwork/memory.md after every meaningful change so later sessions can resume quickly.

Inspired by GitHub projects

This project borrows architecture and practical ideas from the open-source ecosystem:

• OpenCV image stylization and filtering pipelines:
  • https://github.com/opencv/opencv
• Real-time segmentation and media processing design:
  • https://github.com/google-ai-edge/mediapipe
• OBS plugin and streaming workflow patterns:
  • https://github.com/obsproject/obs-studio
• Community camera style transfer experiments:
  • https://github.com/topics/style-transfer

No code is copied from these repositories. References are used for learning design patterns and best practices.

Roadmap

• Add face-aware skin smoothing with landmark protection.
• Add temporal stabilization to reduce frame flicker.
• Add ONNX Runtime backend for model-based stylization.
• Add preset manager and hotkey switching.
• Add benchmark mode for FPS/latency profiling.

License

MIT