July 29, 2026
Speak now or get benchmarked
Show HN: Qwen Scribe – local transcription and dictation for Apple Silicon
A privacy-first voice app lands on Mac — and the comments instantly turn into a speed fight
TLDR: Qwen Scribe is a new Mac app that turns speech into text privately on your own machine instead of sending it to the internet. Commenters immediately made it a showdown, grilling it on speed, language support, and whether it can beat better-known rivals people already trust.
A new Mac-only app called Qwen Scribe just strutted onto Show HN promising something very seductive in 2026 tech terms: your voice stays on your own computer. No account, no cloud, no sending recordings off to some mystery server. It can turn audio and video into text, and even lets you hold the right Command key to speak into almost any text box. Very slick, very private, very "finally, my laptop can mind its own business."
But the real show started in the comments, where the community wasted zero seconds before turning this launch into a full-on leaderboard brawl. One person went straight for the throat with the classic internet challenge: is it actually better than OpenAI's Whisper? Another asked how it stacks up against other fan-favorite tools, while someone else basically walked in holding a stopwatch: "This is like 2x realtime on M1 vs 60x" another app setup, which is the kind of comment that instantly changes the vibe from "cool project" to "receipts, please."
There was also some armchair detective work about why Qwen's speech model isn't showing up in more dictation apps yet, plus the practical question every normal human asks first: what languages does it support? That's the community mood in a nutshell — curious, impressed, but absolutely not ready to hand out a crown. The launch got congratulations, sure, but also the unmistakable Hacker News energy of "nice demo, now survive the comparison chart."
Key Points
- •Qwen Scribe is a local transcription and system-wide dictation app for Apple Silicon Macs that keeps audio and transcript data on-device.
- •It runs Qwen3-ASR through mlx-qwen3-asr on the Mac’s Metal GPU and offers 1.7B and 0.6B model options.
- •The current release is v0.1.0-beta.1 and is source-only; users currently build the app locally with `make app`.
- •Features include drag-and-drop audio/video transcription, automatic language detection, timestamps, SRT export, local transcript history, and right-Command-triggered dictation.
- •The app requires Apple Silicon, macOS 14+, Python 3.10+, ffmpeg for non-WAV/video input, and additional permissions for microphone, input monitoring, and accessibility for dictation.