Lenia Lab

A Tkinter-based Lenia sandbox for building multi-channel cellular automata, tracking organisms, and recording results.

Showcase

1-channel cells (full board)	3-channel cells (full board)
3-channel split views

Requirements

• Python 3.10+ with Tkinter available
• Packages: torch, numpy, pillow, matplotlib, scikit-image, imageio, pandas
• GPU is optional; PyTorch will use CUDA if available

Install (example):

python -m venv .venv
.venv\Scripts\activate
pip install torch numpy pillow matplotlib scikit-image imageio pandas
# For the best PyTorch wheel, see https://pytorch.org/get-started/locally/

Run

python app.py

Data & folders

• presets/ - bundled presets shipped with the app
• saved_settings/ - your saved simulation settings (*.json)
• saved_settings/organisms/ - saved organisms/presets you capture from the board (*.json)
• gifs/ - recordings written by the recorder (GIFs, stats, and snapshots)

The preset list shows everything together with prefixes:

• [Preset] built-in items from presets/
• [Saved Organism] your captured organisms from saved_settings/organisms/
• [Saved Settings] full simulation configs from saved_settings/

Quick usage

• Spacebar toggles pause/play.
• Left-drag draws mass on the selected channel; right-drag erases; Shift+Click selects an organism (when tracking is enabled).
• Add and configure channels in the "Channels & Kernels" tab; adjust interactions via the matrix.
• Use the visualization dropdowns to switch between final board, potential, growth, and flow fields; toggle split-screen or zoom view as needed.

Saving & loading

• Save Settings stores the entire simulation configuration to saved_settings/<name>.json.
• Load Settings restores a saved configuration from that folder.
• Save Selected (in the presets panel) saves the currently selected organism to saved_settings/organisms/<name>.json.
• The list allows loading any entry; built-in presets are read-only, while saved organisms/settings can be renamed or deleted.

Recording

• Record Full GIF captures the rendered board to a GIF under gifs/.
• Record Organism Stats logs per-frame metrics and writes cropped GIFs for multiple view modes into a timestamped folder under gifs/.
• Recording resolution is set via RECORDING_RESOLUTION in config.py (set to None to keep native size).

AI assistance

• Some scripting and refactors were produced with help from Google's Gemini 2.5 Pro.

Notes

• The first run creates needed folders automatically.
• If PyTorch cannot find CUDA, it falls back to CPU automatically.