A Rust HEVC Decoder on Your Home Computer
A Rust HEVC decoder on your home computer: rusty_h265 0.7.0 decodes measured clips 1.37x to 1.50x faster than 0.6. It is host software, and no camera path uses it yet.

Can a Rust HEVC decoder run on my home computer today?
As a library, yes. rusty_h265 0.7.0 is published on crates.io and was measured on a desktop-class x86-64 machine. remade_ffmpeg_rs pull request 98 merged to main at 10:37 AM MST on 7 November 2026, and commit ee52ed9b pins rusty_h265 and rusty_h265-accel 0.7.0. Merge CI was still running at 10:40 AM MST, so this is not a finished tool release. No dldeploy camera path decodes HEVC with it yet.
A Rust HEVC decoder on your home computer is now a little cheaper to run. rusty_h265 0.7.0, published on crates.io on 7 November 2026, decodes three measured clips 1.37x, 1.39x and 1.50x faster than its own 0.6 build. That is host software on a desktop-class chip. It is not firmware, it is not on any ESP32, and no dldeploy camera path uses it yet.
This page is part of the series that starts at smart home access on your home computer. Each page names one job, what already runs on the LAN, and what is still unmeasured. The job here is decoding HEVC video, the format also called H.265, on the machine you own. The format is defined by the ITU-T H.265 recommendation.
It is worth being plain about what is new. The decoder existed before. What changed on 7 November is a measured instruction-count campaign, released as 0.7.0 from commit c01a7b2b. Every number on this page is repository-reported. We reran nothing.
What a Rust HEVC decoder measured on a home computer
The campaign measured one decoder against its earlier self. The instruction counts came from callgrind, the Valgrind counting tool, under WSL on the shipping build with AVX2 kernels. A 20-second 720p x265 clip fell 34.0 percent in instructions. An all-intra clip fell 36.7 percent. A deblocking-heavy clip fell 26.3 percent.
The time comparison ran on an Intel i7-14650HX, a laptop-class x86-64 processor, one pinned core, alternating old and new builds. The new build won 11 of 11 pairs on each clip:
- the 20-second 720p x265 clip, 1.37x
- a 60-frame JCT-VC 720p clip, 1.39x
- the all-intra clip, 1.50x
Those are the project's ratios on the project's machine. A different home computer, with a different CPU and more than one core busy, will see a different number. The measured build used AVX2 kernels, and that matters for anyone comparing an older machine.
The decoder covers HEVC Main, Main 10 and Main Still Picture at 4:2:0, 8 and 10 bit. The campaign reports 147 of 147 JCT-VC conformance streams decoding bit-exact after every change. That is the correctness gate on a fixed test suite, not a guarantee about every file a camera writes. The decoder is written in Rust, which reduces the risk of one class of memory bugs in a parser that reads untrusted files.
What a Rust HEVC decoder on your home computer is not
A good series page says what it is not before anyone has to ask. This one has three boundaries.
A Rust HEVC decoder is not firmware
The boards dldeploy flashes run firmware built on Espressif's ESP-IDF. Nothing in the rusty_h265 0.7.0 campaign runs on an ESP32, an ESP32-S3, or any firmware target. The same media repository has other embedded notes, a plan for H.264 and JPEG on ESP32-class parts and a FLAC encoder run on an ESP32-S3, and those are separate stories. A single-thread AVX2 measurement on a laptop says nothing about what a microcontroller can do.
Not faster than FFmpeg on ordinary clips
The same harness timed FFmpeg 8.1.2 on one thread. The Rust HEVC decoder was 1.37x faster on one weighted-prediction stream, roughly 0.80x to 0.86x as fast on ordinary 8-bit clips, and level on 10-bit and deblocking-heavy clips. On everyday video, FFmpeg's single-thread decoder is still ahead. FFmpeg with its own threading was not part of the comparison at all.
Not the container change
An earlier dldeploy page, our FFmpeg alternative stopped reading the whole file, covered opening recordings without reading the whole file into memory first. That was the container side. This page is the decoder side, the part that turns compressed frames into pictures. They are different changes, and the decode campaign does not change how files are opened.
Where a Rust HEVC decoder on your home computer fits next
The series rule asks for two lists: the access already on the LAN, and the rows still unmeasured. Here they are, then one proposal, clearly labeled.
What already runs on the LAN
The camera page is the measured access. A flashed ESP32 camera serves an MJPEG page that a browser on the same network opens directly, covered in the rusty_esp_video field note. Earlier measurements put that page at about 13.8 to 14.3 frames per second under a 15 fps cap on a XIAO ESP32-S3 Sense. That is previous context, not a new benchmark, and it is MJPEG, not HEVC. The comparison with the stock sketch is a pure Rust alternative to CameraWebServer. The camera crates belong to the Janus family at remadewithrust.com, which is why that site is linked here.
What a Rust HEVC decoder still leaves unmeasured
- The home computer's media controller is not wired to a Janus stream.
- No dldeploy camera or recording path decodes HEVC with rusty_h265.
- remade_ffmpeg_rs pull request 98 merged to main at 10:37 AM MST on 7 November 2026. Commit ee52ed9b pins rusty_h265 and rusty_h265-accel 0.7.0, and the mixed speed sentence is on that readme. Merge CI, portfolio-check, and release-plz were still running at 10:40 AM MST, so this is not a finished tool release. An earlier MP3 and AAC cargo-vet miss, rusty_alloc 2.4 missing safe-to-run, is not an HEVC decode failure.
- Decode time on other CPUs, with more threads, or on a busy machine is not measured.
- Reach from outside the house is not measured.
A proposed workflow, not a built one
Many households already have HEVC files: phones can record it, and Apple documents HEVC capture on its devices. One proposal is a home computer that opens those recordings from its own disk and decodes them with the Rust HEVC decoder, instead of uploading them to a cloud converter. That is a suggestion. No dldeploy page, tool or service does it today. The home computer that would hold those files is MATA, because that is the machine this series is about. The camera vision is See, and the reason a household might want this at all is when a home camera company watches your house.
FAQ
Quick answers for builders evaluating this technology.
Does an ESP32 camera decode or stream HEVC with this release?
No. Nothing in the rusty_h265 0.7.0 campaign runs on an ESP32 or any firmware target. The camera page our boards already serve on the LAN is MJPEG, measured earlier at about 13.8 to 14.3 frames per second under a 15 fps cap.
How much faster is rusty_h265 0.7.0 on a home computer?
The repository reports 1.37x, 1.39x and 1.50x over the 0.6 build on three named clips, one pinned core of an Intel i7-14650HX. Your computer is a different machine, so treat those as the project's ratios, not a promise for yours.
Is this Rust HEVC decoder faster than FFmpeg?
Not overall. Against FFmpeg 8.1.2 on one thread, the repository reports 1.37x faster on a weighted-prediction stream, roughly 0.80x to 0.86x as fast on ordinary 8-bit clips, and parity on 10-bit and deblocking-heavy clips.
Which home access already runs on the LAN, and which is unmeasured?
Already on the LAN: the MJPEG camera page an ESP32 serves to a browser. Unmeasured or unbuilt: the home computer's media controller receiving a Janus stream, any HEVC step in a dldeploy camera path, and reach from outside the house.
What would an HEVC workflow on a home computer look like?
One proposal, not built: a home computer opens HEVC recordings already on its own disk and decodes them with rusty_h265 instead of uploading them to a cloud converter. No dldeploy page, tool or service does this today.