This project generates polished "thought of the day" images for the International Consciousness Research Laboratories (ICRL).
It uses LiteLLM to:

1. Turn a text thought into an image prompt
2. Generate an AI image
3. Blend the generated image with an overlay image

The script is exposed as a command-line tool using Typer and can be run locally or via Docker.

• Generate text-to-image prompts using LiteLLM (completion)
• Produce images using LiteLLM (image_generation)
• Blend generated images with overlay images using configurable alpha values
• Batch process multiple thoughts from CSV files
• Save:
  • the final blended image
  • the raw generated image (optional)
  • the generated image prompt (optional)
• Easy CLI interface
• Docker support with multi-stage builds for minimal image size

pip install icrl-generator

git clone https://github.com/HonakerM/icrl-generator.git
cd icrl-generator
pip install .

Pull the pre-built image from GitHub Container Registry:

docker pull ghcr.io/honakerm/icrl-generator:latest

Or build locally:

docker build -t icrl-generator .

After installation, you can run the following command to generate an image.

Linux:

export OPENAI_API_KEY="<openai-api-key>"
python -m icrl_generator generate-image \
    "Never assume that what you see on a person's face is what lies in their heart." \
    overlay.png \
    output.png \
    --image-prompt-output image-prompt.txt \
    --raw-image-output raw_image.png \
    --log-level DEBUG \
    --alpha 50

Windows:

$env:OPENAI_API_KEY = "<openai-api-key>"   
python -m icrl_generator generate-image `
    "Never assume that what you see on a person's face is what lies in their heart." `
    overlay.png `
    output.png `
    --image-prompt-output image-prompt.txt `
    --raw-image-output raw_image.png `
    --log-level DEBUG `
    --alpha 50

To view all options and their descriptions:

python -m icrl_generator generate-image --help

To generate multiple images from a CSV file:

Linux:

export OPENAI_API_KEY="<openai-api-key>"
python -m icrl_generator generate-batch \
    thoughts.csv \
    overlay.png \
    output_folder \
    --include-image-prompt \
    --include-raw-image \
    --log-level INFO \
    --alpha 50

Windows:

$env:OPENAI_API_KEY = "<openai-api-key>"
python -m icrl_generator generate-batch `
    thoughts.csv `
    overlay.png `
    output_folder `
    --include-image-prompt `
    --include-raw-image `
    --log-level INFO `
    --alpha 50

The CSV file should have thoughts in column 2 (index 1) and publish dates in the second-to-last column in YYYY-MM-DD format.

To view all options:

python -m icrl_generator generate-batch --help

Linux/Mac:

docker run --rm \
  -v $(pwd)/overlay.png:/app/overlays/overlay.png:ro \
  -v $(pwd)/output:/app/output \
  -e OPENAI_API_KEY="sk-your-api-key-here" \
  ghcr.io/honakerm/icrl-generator:latest \
  generate-image \
  "Never assume that what you see on a person's face is what lies in their heart." \
  /app/overlays/overlay.png \
  /app/output/result.png \
  --alpha 50 \
  --log-level INFO

Windows PowerShell:

docker run --rm `
  -v ${PWD}/overlay.png:/app/overlays/overlay.png:ro `
  -v ${PWD}/output:/app/output `
  -e OPENAI_API_KEY="sk-your-api-key-here" `
  ghcr.io/honakerm/icrl-generator:latest `
  generate-image `
  "Never assume that what you see on a person's face is what lies in their heart." `
  /app/overlays/overlay.png `
  /app/output/result.png `
  --alpha 50 `
  --log-level INFO

Linux/Mac:

docker run --rm \
  -v $(pwd)/thoughts.csv:/app/input/thoughts.csv:ro \
  -v $(pwd)/overlay.png:/app/overlays/overlay.png:ro \
  -v $(pwd)/output:/app/output \
  -e OPENAI_API_KEY="sk-your-api-key-here" \
  ghcr.io/honakerm/icrl-generator:latest \
  generate-batch \
  /app/input/thoughts.csv \
  /app/overlays/overlay.png \
  /app/output \
  --alpha 50 \
  --log-level INFO

Windows PowerShell:

docker run --rm `
  -v ${PWD}/thoughts.csv:/app/input/thoughts.csv:ro `
  -v ${PWD}/overlay.png:/app/overlays/overlay.png:ro `
  -v ${PWD}/output:/app/output `
  -e OPENAI_API_KEY="sk-your-api-key-here" `
  ghcr.io/honakerm/icrl-generator:latest `
  generate-batch `
  /app/input/thoughts.csv `
  /app/overlays/overlay.png `
  /app/output `
  --alpha 50 `
  --log-level INFO

If you want to call it from Python directly:

from pathlib import Path
from icrl_generator import generate_image

generate_image(
    thought="Reality responds to our expectations.",
    overlay=Path("overlay.png"),
    output=Path("out.png"),
    image_prompt_output=Path("prompt.txt"),
    raw_image_output=Path("raw.png"),
    alpha=50
)

The --alpha parameter controls the blend between the generated image and the overlay:

• 0: Fully generated image (no overlay)
• 50: Equal blend (default)
• 100: Fully overlay image (no generated image visible)

When using Docker, you need to mount directories and files:

Before running Docker commands:

1. Create the output directory: mkdir -p output
2. Place your overlay.png in the current directory
3. For batch processing, place your thoughts.csv in the current directory

• Python 3.10+
• LiteLLM API keys properly configured (export OPENAI_API_KEY=...)
• Pillow
• Typer
• tqdm (for batch processing progress bars)

Generate a single image with overlay blending.

Arguments:

• THOUGHT: The thought text to generate an image for
• OVERLAY: Path to overlay image to blend with generated image
• OUTPUT: Path to save the final blended image

Options:

• --image-prompt-output PATH: Save the generated image prompt
• --raw-image-output PATH: Save the raw image before blending
• --log-level: Set logging level (default: INFO)
• --prompt-gen-model TEXT: LLM model for prompt generation (default: gpt-4o-mini)
• --image-gen-model TEXT: Image generation model (default: gpt-image-1-mini)
• --alpha INTEGER: Blend alpha value 0-100 (default: 50)

Generate multiple images from a CSV file.

Arguments:

• INPUT: Path to CSV file with thoughts and publish dates
• OVERLAY: Path to overlay image to blend with all generated images
• OUTPUT_FOLDER: Folder to save generated images

Options:

• --include-image-prompt/--no-include-image-prompt: Save image prompts (default: no)
• --include-raw-image/--no-include-raw-image: Save raw images (default: no)
• --log-level: Set logging level (default: INFO)
• --prompt-gen-model TEXT: LLM model for prompt generation (default: gpt-4o-mini)
• --image-gen-model TEXT: Image generation model (default: gpt-image-1-mini)
• --alpha INTEGER: Blend alpha value 0-100 (default: 50)

docker build -t icrl-generator .

The Dockerfile uses multi-stage builds to create a minimal runtime image by:

1. Building dependencies in a builder stage with compilation tools
2. Copying only the runtime files to a slim final image
3. Excluding build tools to minimize image size

This project uses GitHub Actions to automatically build and publish Docker images to GitHub Container Registry (GHCR) on:

• Pushes to main or develop branches
• Version tags (e.g., v1.0.0)
• Pull requests (build only, no push)

Images are published to: ghcr.io/honakerm/icrl-generator