Extensions/ComfyUI-Image-Compress
ComfyUI Extension

ComfyUI-Image-Compress

A custom ComfyUI node that compresses images using JPEG or PNG encoding with independent control over encode quality and output scale. Includes a detailed per-frame performance report covering file size savings and timing metrics.

By comfyuiattic-989·Created 5 months ago·Updated 5 months ago· 1
comfyuiattic-989/comfyui-image-compress
Nodes1
On cloudLocal install
Categoryimage/processing
Stars1
Updated5 months ago

Nodes (1)

Image Compress
image/processing
Readme

ComfyUI Image Compress

CI License: MIT Python 3.10+ ComfyUI

A custom ComfyUI node that compresses images using JPEG or PNG encoding with independent control over encode quality and output scale. Includes a detailed per-frame performance report covering file size savings and encode/decode timing.

ko-fi

Features

  • JPEG lossy compression — quality slider directly degrades pixel data, exactly as it would appear saved to disk
  • PNG lossless compression — deflate level controls file size on disk; pixels are always preserved exactly
  • Scale percent — resizes image dimensions before encoding; the only way to visibly reduce a PNG output tensor
  • Batch support — processes all frames in a batch tensor independently
  • Performance report — per-frame metrics for original size, compressed size, reduction %, compression ratio, PNG deflate level, encode time, and decode time; batch summary included for multi-frame inputs
  • Round-trip encoding — output tensor reflects the actual encoded result, so JPEG artefacts are baked in downstream

Installation

Option A — Clone directly into custom_nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/comfyuiattic-989/comfyui-image-compress

Option B — Manual:

  1. Download the latest release zip from the Releases page
  2. Extract the image_compress folder into ComfyUI/custom_nodes/

Then restart ComfyUI. The node will appear under image/processing → Image Compress.

No extra dependencies required. Pillow, NumPy, and PyTorch are all bundled with ComfyUI.

Usage

Inputs

| Input | Type | Default | Description | |---|---|---|---| | image | IMAGE | — | Standard ComfyUI image tensor. Batches are fully supported. | | quality | INT slider 1–100 | 80 | JPEG: encode quality — lower = more artefacts, smaller file. PNG: deflate level — lower = more compression, same pixels. | | scale_percent | INT slider 1–100 | 100 | Resize both dimensions to this percentage before encoding. 100 = no resize. 50 = half width and height. Applies to both formats. | | format | JPEG | PNG | JPEG | Codec to use for compression. |

Outputs

| Output | Type | Description | |---|---|---| | image | IMAGE | Compressed (and optionally rescaled) image tensor | | report | STRING | Plain-text performance report. Wire to a Show Text node to display in-graph, or read from the server console. |

Understanding the Two Controls

Example-01-Settings.png

The quality and scale_percent sliders are independent and serve different purposes:

| | quality | scale_percent | |---|---|---| | JPEG | Degrades pixel data (lossy) | Reduces pixel dimensions | | PNG | Controls deflate level (file size only — pixels unchanged) | Reduces pixel dimensions |

PNG note: Because PNG is a lossless format, the decoded output tensor will always be pixel-identical to the input regardless of quality. If you need the output tensor to visibly differ when using PNG, set scale_percent below 100. The report will warn you if PNG is run at full scale.

Example combinations

| Format | Quality | Scale | Effect | |---|---|---|---| | JPEG | 80 | 100 | Lossy encode, original dimensions | | JPEG | 40 | 50 | Aggressive lossy encode at half resolution | | PNG | 50 | 100 | Lossless pixels, moderate deflate — tensor unchanged | | PNG | 50 | 75 | Lossless pixels at 75% resolution — tensor visibly smaller |

Sample Report Output

╔════════════════════════════════════════════════════════╗
║  IMAGE COMPRESS – PERFORMANCE REPORT                   ║
╚════════════════════════════════════════════════════════╝
  Format        : JPEG
  Quality       : 75/100
  Scale         : 50%
  Frames        : 1
──────────────────────────────────────────────────────────
  Frame   1
    Input  dimensions:  1024×1024 px
    Output dimensions:   512×512 px  (scaled to 50%)
    Original size    :   3072.00 KB  (uncompressed)
    Compressed size  :     48.21 KB
    Size reduction   :     98.4 %
    Compression ratio:   63.72×
    Encode time      :      6.38 ms
    Decode time      :      2.11 ms
    Total time       :      8.49 ms
──────────────────────────────────────────────────────────

Project Structure

comfyui-image-compress/
├── .github/
│   ├── ISSUE_TEMPLATE/
│   │   ├── bug_report.md
│   │   ├── feature_request.md
│   │   └── config.yml
│   └── workflows/
│       └── ci.yml          # Lint + smoke tests on Python 3.10/3.11/3.12
├── __init__.py             # ComfyUI node registry entry point
├── image_compress.py       # Node implementation
├── pyproject.toml          # Package metadata
├── CHANGELOG.md
├── LICENSE                 # MIT
└── README.md

Requirements

  • ComfyUI (any recent version)
  • Python 3.10+
  • Pillow >= 9.0 — bundled with ComfyUI
  • numpy >= 1.23 — bundled with ComfyUI
  • torch >= 2.0 — bundled with ComfyUI

Contributing

Bug reports and feature requests are welcome via GitHub Issues. Please use the provided issue templates.

For pull requests:

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/my-change)
  3. Make your changes
  4. Ensure ruff check . passes with no errors
  5. Open a pull request against main

Changelog

See CHANGELOG.md for the full version history.

License

MIT © 2026

ko-fi