image_forge_core

Open-source project maintained by the community. Found a bug or want to contribute? PRs and issues are welcome.

Lightweight Rust image processing engine for Flutter — core operations only. GPU-accelerated resize, crop, rotate, compress, thumbnails, basic filters, and multi-format encoding. Runs native Rust on device with SIMD and WGSL compute shaders.

Platform Support

Platform	Status
Android	Tested (API 21+)
iOS	Tested (15.0+)
macOS	Tested (12+)
Windows	In progress
Linux	In progress
Web	Not supported

Quick Start

import 'package:image_forge_core/image_forge_core.dart';

// Resize an image
final thumb = resizeImage(
  bytes: bytes,
  width: 512,
  height: 512,
  algorithm: ResizeAlgorithm.lanczos3,
  format: OutputFormat.jpeg,
  quality: 85,
  fixExif: true,
  backend: ProcessingBackend.auto,
);

// Compress an image
final compressed = compressImage(
  bytes: bytes,
  format: OutputFormat.jpeg,
  quality: 70,
);

// Create a thumbnail (fits within max_edge bounding box)
final thumb = createThumbnail(
  bytes: bytes,
  maxEdge: 512,
  format: OutputFormat.jpeg,
  quality: 85,
  algorithm: ResizeAlgorithm.lanczos3,
  fixExif: true,
  backend: ProcessingBackend.auto,
);

// Probe image info without full decode
final info = probeImage(bytes: bytes);
print('${info.width}x${info.height}');

What You Get

Resize, crop, rotate — Multiple resize algorithms (Lanczos3, Mitchell, CatmullRom, etc.) with batch parallelism.
Compress & thumbnails — MozJPEG and oxipng optimization, fast thumbnail generation.
Basic filters — Brightness, contrast, saturation, hue, blur, sharpen, warmth, fade, vignette, highlights, shadows, structure, pixelize, frosted glass, oil, solarize, and 14 classic presets.
Drawing & overlays — Text, lines, circles, watermark image overlays with blend modes (Multiply, Screen, Overlay, Add).
EXIF handling — Read and fix orientation without re-encoding.
Multi-format — JPEG (MozJPEG), PNG (oxipng), WebP, AVIF encode/decode.
Progressive decode — Low-res preview + full-res buffer in one pass.
GPU compute — wgpu Metal/Vulkan acceleration for resize, blur, sharpen, and color adjustments.
Buffer pool — Zero-allocation byte vector reuse for render loops.
BlurHash — Placeholder generation for loading states.

For the full API, see the Dart API reference.

App Size

The package bundles native Rust libraries per platform. Size impact by component:

Component	Est. Size	What you lose if removed
Core engine (resize, crop, rotate, compress, EXIF, filters, drawing)	~7-9 MB	—
GPU compute (wgpu — Metal/Vulkan)	+4-6 MB	Hardware-accelerated resize, blur, sharpen, brightness/contrast/saturation/hue
AVIF encode/decode (rav1e)	+1-2 MB	AVIF format support

Installation

flutter pub add image_forge_core

Prerequisites: A Rust toolchain on your development machine. For Android:

rustup target add aarch64-linux-android armv7-linux-androideabi \
  x86_64-linux-android i686-linux-android

More Examples

Crop & rotate

final cropped = cropImage(
  bytes: bytes, x: 100, y: 100, width: 400, height: 300,
  format: OutputFormat.jpeg, quality: 85, fixExif: true,
);

final rotated = rotateImage(
  bytes: bytes, rotation: Rotation.rotate90,
  format: OutputFormat.jpeg, quality: 85, fixExif: true,
);

Apply a filter

final filtered = applyFilter(
  bytes: bytes,
  filter: ImageFilter.brightness(amount: 25),
  format: OutputFormat.jpeg,
  quality: 85,
  fixExif: true,
);

Overlay with blend mode

final composed = overlayImage(
  baseBytes: photo, overlayBytes: logo,
  x: 40, y: 40, blendMode: BlendMode.multiply,
  format: OutputFormat.png, quality: 100,
);

Progressive decode (low-res preview + full buffer)

final result = decodeProgressiveImage(
  bytes: bytes,
  previewMaxEdge: 128,
  fixExif: true,
);
// Show result.previewRgba immediately for instant feedback
// Then work with result.buffer for full-resolution editing

Batch resize (parallel)

final results = batchResizeImages(
  items: [
    BatchResizeItem(bytes: img1, width: 512, height: 512),
    BatchResizeItem(bytes: img2, width: 256, height: 256),
  ],
  algorithm: ResizeAlgorithm.lanczos3,
  format: OutputFormat.jpeg,
  quality: 85,
  backend: ProcessingBackend.auto,
);

GPU detection

final gpu = gpuComputeInfo();
if (gpu.available) print('GPU: ${gpu.device} (${gpu.api})');

final thumb = resizeImage(
  bytes: bytes, width: 512, height: 512,
  algorithm: ResizeAlgorithm.lanczos3,
  format: OutputFormat.jpeg, quality: 85,
  fixExif: true,
  backend: ProcessingBackend.auto, // auto = GPU if available
);

RGBA buffer pipeline (zero intermediate encode/decode)

final buf = decodeToRgbaBuffer(bytes: bytes, fixExif: true);

// Chain operations on raw pixels
final filtered = filterRgbaBuffer(
  buffer: buf,
  filter: ImageFilter.blur(radius: 4),
  backend: ProcessingBackend.auto,
);
final resized = resizeRgbaBuffer(
  buffer: filtered,
  width: 1024, height: 768,
  algorithm: ResizeAlgorithm.lanczos3,
  backend: ProcessingBackend.gpu,
);

// Encode once at the end
final out = encodeRgbaBuffer(
  buffer: resized,
  format: OutputFormat.jpeg,
  quality: 90,
);

Contributing

This package is part of the MediaForge monorepo. Issues and pull requests are welcome on GitHub.