Release
OpenASR Desktop 0.1.21: Voice ID
Voice ID lands in Desktop 0.1.21. Enroll a short sample, keep speakers apart with MOSS, and — when you choose — put real names on the transcript. Audio stays on your device.
Blog
Notes from the local-first frontier.
Release notes, engineering deep-dives, and field reports on running speech recognition entirely on your own hardware.
Release 0.1.x in ten days: what fourteen versions built
Fourteen versions in ten days. The 0.1.x series was a foundation-laying sprint — halving resident memory, shipping 26 models across eleven platform targets, and hunting every fail-closed gap until idle unload actually worked. Here's what it built.
Release OpenASR Desktop is now on Windows
Same local-first speech-to-text, now on Windows. Same open core, same privacy boundary, same models. Here's what to expect — including the SmartScreen prompt you'll see on first launch.
Release OpenASR 0.1.0
The first public release: a local speech-to-text app for macOS over an Apache-2.0 open core. Here's what works today, what's still missing, and where to read the source.
Deep dive Who-said-what, the dependency-light way: a pure-Rust late-fusion diarization substrate
VAD and diarization run as late-fusion sibling stages around an untouched ASR core, talking only through time intervals — the project's first pure-Rust forward passes, validated bit-close to their ONNX references.
Deep dive Trusting a model you didn't compile: a signed catalog and a fail-closed pull chain
The .oasr pack is pinned, the catalog is signed, and nothing is auto-downloaded — a walk through OpenASR's fail-closed pull chain and where its trust root actually lives.
Engineering Batched serve-mode decode: GPU-only throughput that leaves single-stream byte-identical
A default-off OPENASR_SERVE_BATCH path batches concurrent GPU decode into one GEMM per step. Measured 4.28x at N=8 on one GPU with moonshine-tiny — and N=1 stays byte-identical to today's path.
Deep dive Moving the KV cache onto the GPU: turning a transfer-bound decoder compute-bound
A profile of one decode on one AMD GPU put the bottleneck on the host-to-GPU KV round-trip, not on compute. Making the KV cache device-resident cut the per-token KV/IO from 80.6 to 1.1 ms/token — scoped to that one GPU and model.
Engineering Benchmarking OpenASR against whisper.cpp: one host, one clip, an honest number
How the bench-suite harness measures OpenASR against whisper.cpp on the same model — a committed 7.9% on one host, an honest spread, and a regression guard rather than a standing OSS win.
Deep dive Inside the .oasr format: one file for weights, tokenizer, and metadata
Why a model is a single GGUF-backed .oasr file instead of a directory of loose tensors - and why that pack is pinned and catalog-signed, not signed in place.
Privacy Local by default: privacy you can actually audit
Audio is transcribed where it's recorded. A walk through the offline-default, no-auto-download boundary — from explicit openasr pull, to a loopback-only API, to anonymous diarization.
Engineering Drop-in compatible: matching the OpenAI transcription API on localhost
How openasr serve mirrors a focused slice of the OpenAI transcription HTTP shape on loopback — so you switch a base URL, not your client — while staying honest that it is a subset, not the full API.
Deep dive What quantization actually trades: size, speed, and the long tail of quality
fp16, q8_0, q4_k, and Qwen's q3_k on a single clip: where compression is almost free, where a size win isn't a speed win, and why moonshine-tiny's 11.76% WER is the model, not the quantizer.