GLM Image Metadata Extractor - Python Implementation

This Python implementation was inspired by glmimage.app. It provides efficient extraction and analysis of metadata from AI-generated images, including model signatures, generation parameters, and provenance tracking.

Features

• PNG Chunk Parsing: Pure Python implementation for reading PNG chunks
• EXIF Metadata Extraction: Extract standard EXIF data from images
• AI Signature Validation: Verify and validate AI model signatures
• Provenance Tracking: Track the origin and generation parameters of AI images
• Type Hints: Full type annotations for better IDE support
• Dataclasses: Modern Python dataclass-based API

Installation

cd python
pip install -r requirements.txt

Development Dependencies

pip install -e ".[dev]"

Usage

Basic Usage

from main import extract_ai_metadata, print_metadata_summary

# Extract metadata from an image
metadata = extract_ai_metadata("generated-image.png")
print_metadata_summary(metadata)

Creating Metadata Programmatically

from main import create_metadata_from_data, validate_ai_signature

metadata = create_metadata_from_data({
    "model": "GLM-4V-Image",
    "prompt": "A beautiful sunset over mountains",
    "parameters": {"steps": "50", "guidance": "7.5"},
    "width": 1024,
    "height": 1024,
    "format": "PNG"
})

is_valid = validate_ai_signature(metadata)
print(f"Valid: {is_valid}")

Running the Demo

python main.py

API Reference

extract_ai_metadata(image_path: str) -> AIMetadata

Extracts AI metadata from an image file. Returns an AIMetadata dataclass.

validate_ai_signature(metadata: AIMetadata) -> bool

Validates the AI model signature in the metadata. Returns True if valid.

embed_provenance_metadata(metadata: AIMetadata, source: str) -> None

Adds provenance information to the metadata object.

print_metadata_summary(metadata: AIMetadata) -> None

Prints a formatted summary of the extracted metadata to the console.

create_metadata_from_data(data: dict) -> AIMetadata

Creates a new AIMetadata instance from a dictionary.

Data Structures

@dataclass
class AIMetadata:
    model: str = ""
    prompt: str = ""
    parameters: Dict[str, str] = field(default_factory=dict)
    timestamp: str = ""
    provenance: str = ""
    signature: str = ""
    width: int = 0
    height: int = 0
    format: str = ""

Testing

pytest

Code Formatting

black main.py

Type Checking

mypy main.py

Links

• Source: glmimage.app
• Repository: https://github.com/user/glmimage-app-oss
• PyPI Package: https://pypi.org/project/glmimage-app-oss

License

MIT License - see LICENSE for details.