yscv-video 0.1.7

Video decoding (H.264, HEVC), MP4 parsing, and camera I/O
Documentation

yscv-video

H.264 and HEVC video decoders, MP4/MKV container parsers, hardware decode, and camera input. Pure Rust, no FFmpeg.

use yscv_video::Mp4VideoReader;

let mut reader = Mp4VideoReader::open("video.mp4")?;
while let Some(frame) = reader.next_frame()? {
    // frame.rgb8_data: Vec<u8>, frame.width, frame.height
}

Capabilities

Feature Details
H.264 decode CAVLC, CABAC, I/P/B slices, 4x4/8x8 DCT, weighted prediction
HEVC decode CTU quad-tree, branchless CABAC, 8-tap MC, SAO, deblock
MP4 parse Streaming moov-only reader, O(1) memory per frame
MKV parse EBML demuxer with frame index
HW decode VideoToolbox (macOS), VA-API (Linux), NVDEC, MediaFoundation
Camera Live capture via Nokhwa (optional)
Audio AAC/ALAC/Opus/Vorbis/MP3/FLAC codec detection
Color YUV420→RGB8 with NEON/SSE2 SIMD

Performance

  • H.264 decode: 4.5x faster than ffmpeg (1080p, Apple M-series)
  • HEVC decode: 1.3x faster than ffmpeg (P/B frames)
  • 29 NEON + 31 SSE2 SIMD functions with runtime dispatch

Features

[features]
videotoolbox = []      # macOS HW decode
vaapi = []             # Linux HW decode
nvdec = []             # NVIDIA HW decode
media-foundation = []  # Windows HW decode
native-camera = []     # Live camera input

Tests

220 tests covering decode correctness, container parsing, edge cases.