pito-codecs

A terminal running pcod on a 1080p H.264 clip: it probes the file, decodes it on the CPU and on the GPU through VA-API with the speed of each, then encodes it to HEVC on the GPU.

Video and audio codecs in pure Rust, on the CPU and the GPU, with no ffmpeg inside. H.264, H.265, VP9, AAC, AC-3, E-AC-3, Opus and MP3; MP4, MOV, MKV and WebM; GIF and WebP. One MIT Cargo package, a feature per part, two direct dependencies and no C linked.

It comes as a Rust library and as pcod, a small command line over it that probes, decodes and encodes MP4 files.

First frame 7× to 21× faster than ffmpeg: 4.1 ms against 87.4 ms for a 1440p H.264 file on the CPU.
Filmstrips 8× to 12× faster: 12 stills at 240 px in 29 to 72 ms.
Frame-accurate smart cuts 1.7× to 2.9× faster than an ffmpeg re-encode: only the frames around each cut are encoded.
GPU decode and encode on VA-API 1.1× to 1.3× faster than ffmpeg’s, and frames can stay on the card from decode to encode.
Overlays composed on the card: a marked or captioned export uses 18× to 23× less CPU time than decoding and blending on the CPU.
Small: a 3.9 MB command line, and a library that pulls in 8 crates on Linux.

pito-codecs against ffmpeg 9.0.1, times faster or slower per job

Every speed number here was measured on one machine (AMD Ryzen 9 9950X, AMD Radeon RX 9060 XT on Mesa’s radeonsi VA-API driver, Linux, ffmpeg 9.0.1) on public-domain NASA footage. The repository’s README gives the input, the commands and the load for each.

Where to start

Install pcod or add the library to a Rust project.
Probe, decode and encode with pcod from a terminal.
Use the library: the media front picks the fastest route for each job and says why when it falls back.
First frames and filmstrips, audio as PCM, smart-cut exports, overlays and captions and exports of your own frames.
GPU routes per vendor and system: what is tested and what is built but untested.
Low-level codecs: H.264 decode and encode, GIF and WebP without the front.
The command reference, the formats and the API reference.

Licence

The code is MIT licensed, © Catalin Ilinca. The PITO MD name and its logos are © Catalin Ilinca, all rights reserved, and are not covered by the MIT licence. pito-codecs is the media layer of PITO MD; its source is at github.com/gmrdad82/pito-codecs.