BlazeDiff — a high-performance, pixel-perfect image comparison library for JavaScript. 1.5x faster than pixelmatch, with identical accuracy and output quality. It uses an innovative block-based algorithm to achieve blazing-fast pixel-by-pixel image diffing. Built on the foundation of pixelmatch but with significant optimizations, it's designed for visual testing, CI/CD pipelines, and any application requiring fast, accurate image comparison.
100% API and Result Compatible: BlazeDiff is fully compatible with pixelmatch's API and produces identical results when using the YIQ color space flag.
BlazeDiff is 1.5x faster than pixelmatch while maintaining the same accuracy and output quality. The performance improvement comes from:
- Block-based algorithm: First pass creates dynamic-sized blocks and only processes changed blocks
- Early exit optimization: Returns immediately if no buffer differences are detected (Native C++ in Node)
- 32-bit integer comparisons: Leverages CPU vectorization for faster pixel matching
- JPEG and WebP support: When using the binary with the Sharp transformer, BlazeDiff supports PNG, JPEG and WebP inputs.
| Expected | Actual | Diff |
|---|---|---|
 |
 |
 |
BlazeDiff is available in multiple packages to suit different use cases:
-
@blazediff/core - Core JavaScript library for pixel-by-pixel image comparison. To use the core library in your JavaScript/TypeScript project, see the installation and usage instructions.
-
@blazediff/bin - Command-line interface for comparing images directly from the terminal. To use BlazeDiff from the command line, see the CLI installation and usage guide.
-
@blazediff/wasm - WebAssembly implementation with SIMD optimization for enhanced performance. To use the WASM version for maximum performance, see the WASM package documentation.
50 iterations (3 warmup)
~50% performance improvement on average.
| Benchmark | Pixelmatch | BlazeDiff | Time Saved | % Improvement |
|---|---|---|---|---|
| 4k/1 | 294.86ms | 274.29ms | 20.57ms | 7.0% |
| 4k/1 (identical) | 19.74ms | 2.33ms | 17.41ms | 88.2% |
| 4k/2 | 295.46ms | 274.97ms | 20.49ms | 6.9% |
| 4k/2 (identical) | 22.10ms | 2.59ms | 19.51ms | 88.3% |
| 4k/3 | 373.39ms | 344.06ms | 29.32ms | 7.9% |
| 4k/3 (identical) | 26.61ms | 3.11ms | 23.51ms | 88.3% |
| page/1 | 321.33ms | 92.50ms | 228.83ms | 71.2% |
| page/1 (identical) | 66.57ms | 7.66ms | 58.91ms | 88.5% |
| page/2 | 418.13ms | 375.39ms | 42.75ms | 10.2% |
| page/2 (identical) | 45.99ms | 5.42ms | 40.57ms | 88.2% |
| pixelmatch/1 | 0.86ms | 0.52ms | 0.34ms | 39.7% |
| pixelmatch/1 (identical) | 0.15ms | 0.01ms | 0.13ms | 90.5% |
| pixelmatch/2 | 2.08ms | 2.01ms | 0.07ms | 3.4% |
| pixelmatch/2 (identical) | 0.07ms | 0.01ms | 0.07ms | 89.8% |
| pixelmatch/3 | 0.77ms | 0.31ms | 0.46ms | 59.6% |
| pixelmatch/3 (identical) | 0.15ms | 0.01ms | 0.13ms | 90.5% |
| pixelmatch/4 | 4.15ms | 3.93ms | 0.22ms | 5.3% |
| pixelmatch/4 (identical) | 0.20ms | 0.02ms | 0.18ms | 90.7% |
| pixelmatch/5 | 0.24ms | 0.14ms | 0.11ms | 43.6% |
| pixelmatch/5 (identical) | 0.07ms | 0.01ms | 0.06ms | 89.1% |
| pixelmatch/6 | 0.92ms | 0.89ms | 0.03ms | 2.9% |
| pixelmatch/6 (identical) | 0.07ms | 0.01ms | 0.06ms | 90.1% |
| pixelmatch/7 | 1.90ms | 1.35ms | 0.55ms | 29.1% |
| pixelmatch/7 (identical) | 0.28ms | 0.03ms | 0.25ms | 90.6% |
Benchmarks run on MacBook Pro M1 Max, Node.js 22
50 iterations (3 warmup)
~60% performance improvement on average.
| Benchmark | Pixelmatch | BlazeDiff | Time Saved | % Improvement |
|---|---|---|---|---|
| 4k/1 | 1832.83ms | 818.00ms | 1014.84ms | 55.4% |
| 4k/1 (identical) | 1480.47ms | 499.92ms | 980.55ms | 66.2% |
| 4k/2 | 1927.86ms | 856.67ms | 1071.18ms | 55.6% |
| 4k/2 (identical) | 1589.19ms | 565.47ms | 1023.72ms | 64.4% |
| 4k/3 | 2071.07ms | 927.41ms | 1143.66ms | 55.2% |
| 4k/3 (identical) | 1707.19ms | 574.57ms | 1132.62ms | 66.3% |
| page/1 | 2420.72ms | 737.27ms | 1683.44ms | 69.5% |
| page/1 (identical) | 2169.74ms | 547.40ms | 1622.34ms | 74.8% |
| page/2 | 1866.68ms | 750.66ms | 1116.02ms | 59.8% |
| page/2 (identical) | 1490.23ms | 362.18ms | 1128.05ms | 75.7% |
Benchmarks run on MacBook Pro M1 Max, Node.js 22
Contributions are welcome! Please see the Contributing Guide for details.
MIT License - see LICENSE file for details.
- Built on the excellent pixelmatch algorithm
Built for high-performance image comparison