pixeloe

What it is and what it does

PixelOE is a Python library that transforms photographs and high-resolution images into pixel art using classical image processing rather than machine learning. It works by first expanding the outlines of important visual features based on local contrast, then intelligently downsampling the result to a target resolution. The library offers both a legacy numpy-based API and a modern PyTorch implementation; the torch version supports GPU acceleration and can run CPU-only if needed.

The package provides three interfaces: a Python API for direct integration, a command-line tool for batch processing, and a ComfyUI custom node for integration into node-based workflows. It includes multiple downsampling modes (center, contrast, k-centroid, bicubic, nearest) and optional color palette quantization for classic pixel art aesthetics. The contrast-aware approach preserves fine details and sharp edges that naive downsampling would lose.

Use it for:

• Convert photographs into retro-style pixel art for game assets or artistic projects without training neural networks.
• Batch process image collections via CLI to generate consistent pixel art with configurable detail levels and color palettes.
• Integrate pixelization into ComfyUI workflows as a post-processing or artistic effect node alongside other image operations.
• Preserve fine details in downsampled images for thumbnail generation or low-resolution display while maintaining visual clarity.
• Apply outline expansion alone to improve existing pixelization methods, including neural network-based approaches.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Converts standard images into pixel art using contrast-aware outline expansion and intelligent downsampling, without requiring neural networks or AI models.

Yes, with conditions. Install if you need deterministic, parameter-tunable pixel art generation without AI overhead and can accept the five-dependency footprint (especially torch). The active maintenance, permissive license, and no known vulnerabilities support adoption. Skip if you want a lightweight solution or prefer neural network-based pixelization; the high install friction and torch dependency are real costs for simpler use cases.

Install

pixeloe on PyPI

pip install pixeloe

uv add pixeloe

poetry add pixeloe

Installing pixeloe

Before you install

High install friction due to five runtime dependencies (opencv-python, numpy, pillow, torch, kornia), though the package supports GPU-free operation for core logic. Maintenance is active with recent commits and steady releases.

License in practice

Apache License 2.0 is permissive, allowing commercial and private use with minimal restrictions; attribution required but no copyleft obligations.

Quickstart

pip install pixeloe

import torch
from pixeloe.torch.pixelize import pixelize
from pixeloe.torch.utils import to_numpy, pre_resize

img = pre_resize(img, target_size=256, patch_size=4).cuda().half()
result = pixelize(img, pixel_size=4, thickness=3)
result_img = to_numpy(result)[0]

Requires Python >=3.10; torch and kornia add substantial disk/memory overhead; GPU optional but recommended for performance.

Verify before relying

• Whether the legacy numpy/cv2 API remains stable or will be removed as indicated by deprecation notice.
• Actual performance on hardware other than RTX4090 and whether performance claims apply to typical consumer GPUs.
• Whether kornia is used in the core torch implementation or only in optional utilities.
• Input format requirements and supported image types beyond the webp examples shown.

Package facts

Evidence: pixeloe-0.1.4.tar.gz

Tags

More Graphics packages

Further reading