img2webp-rs
Auto
CANVAS BACKGROUND
ENGINE
Transparency Showcase
Benchmarks of 9 distinct encoding strategies including temporal sequence manipulation.

Standard

Industry standard behavior. Optimal balance, but may discard RGB data in Alpha=0 pixels.

1.31 MB
Default

Pixel Perfect

Ensures RGB integrity in Alpha=0 pixels. Critical for clean edges during scaling.

1.40 MB
-lossless -exact

Backward Sequence

Last frame first, first frame last. Reverses the entire temporal flow.

1.31 MB
-reverse

Ping-Pong Loop

Forward then backward. Eliminates the "jump" artifact in infinite loops.

2.58 MB
-pingpong

Optimized Small

Intelligent mixed-mode strategy with 50% alpha quality reduction. Best for web.

203 KB
-mixed -alpha_q 50

Near Lossless

Perceptual preprocessing that yielding significant bitstream compression.

895 KB
-near_lossless 60

Extreme Slow

Maximum exhaustive search (M6) with best predictive spatial alpha filtering.

1.38 MB
-m 6 -alpha_filter 2

Lossy Sharp

High-contrast preservation optimized via the Sharp YUV conversion pipeline.

209 KB
-lossy -q 75 -sharp_yuv

Lossy Low

Absolute footprint floor for bandwidth constraints. Noticeable artifacts.

125 KB
-q 30 -alpha_q 30
Technical Architecture Deep Dive

Modular Workspace & FFI Layer

The img2webp-rs architecture separates concerns into a decoupled multi-crate workspace:

  • webp_anim (Library): A safe Rust abstraction over libwebp-sys. It manages the full lifecycle of C-allocated muxer objects using Rust's RAII patterns and automated memory cleanup via the Drop trait.
  • img2webp (CLI Driver): A stateful command-line interface featuring a non-linear two-pass argument parser. This enables the ability to interleave frame-specific configurations (quality, duration) with sequence inputs.

Temporal Manipulation Engine

A primary enhancement in img2webp-rs is the Stateful Job Queue. Unlike standard C converters, we buffer the entire encoding plan, allowing flags like -reverse and -pingpong to manipulate the temporal sequence before the expensive encoding logic begins.

Advanced Alpha Filtering & Pipeline

Asset generation follows a rigid 4-stage deterministic pipeline:

  1. Safe Decoding: Imagery is decoded into raw RGBA8 buffers using audited Rust decoders (PNG, TIFF, BMP, etc.).
  2. Minimal Differencing: The WebPAnimEncoder calculates the optimal sub-rectangle delta between sequential frames to minimize bitstream weight.
  3. Pixel-Perfect Integrity: The -exact flag preserves raw RGB data at 0% Alpha, preventing the "Dark Halo" effect common in browser rendering.
  4. Remuxing Engine: Post-assembly metadata injection via WebPMux.
// img2webp-rs Internal Alpha Selection match config.alpha_filtering { 0 => "None: Direct bitstream storage", 1 => "Fast: Horizontal/Vertical heuristic prediction", 2 => "Best: Full spatial neighborhood analysis (M6)", }