--- title: "GitHub - VladUZH/qwen-scribe: Private, local transcription and system-wide dictation for Apple Silicon." url: https://daily.dev/posts/github---vladuzh-qwen-scribe-private-local-transcription-and-system-wide-dictation-for-apple-silic-23jfjb1mk source_url: https://github.com/VladUZH/qwen-scribe type: article source: "Hacker News" published: 2026-07-29T18:35:15.755Z updated: 2026-07-29T19:19:07.906Z tags: ["python", "mac", "local-first"] reading_time: 6 upvotes: 0 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # GitHub - VladUZH/qwen-scribe: Private, local transcription and system-wide dictation for Apple Silicon. **[Hacker News](https://daily.dev/sources/hn)** · 6 min read · 0 upvotes · 0 comments ## Summary Qwen Scribe is an open-source macOS app for Apple Silicon that provides private, fully on-device audio and video transcription using the Qwen3-ASR model via MLX. It supports drag-and-drop file transcription, automatic language detection, SRT export, transcript history, and system-wide dictation triggered by holding the right Command key. No cloud service, API key, or account is required. Currently at v0.1.0-beta.1, it requires building from source. The architecture uses a FastAPI job queue serving a browser UI and a native macOS dictation helper, with GPU jobs serialized to avoid memory contention. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Community take How the wider developer community reacted, aggregated from 1 discussion and 28 comments across hackernews (as of 2026-07-29). **TL;DR:** The community is cautiously positive about Qwen Scribe, appreciating its local-only, privacy-first approach and the Qwen3-ASR model's noise robustness and punctuation quality, while noting it's less mature than alternatives like Handy and significantly slower than Parakeet-based tools. **Sentiment:** 55% positive · 35% mixed · 10% skeptical **The case for** - Qwen3-ASR reportedly handles noisy environments better than Whisper and punctuates more naturally. - Fully local and Apache-2.0 licensed, with no cloud dependency or account required. - Supports vocabulary hints for names and jargon, a practical advantage over Whisper/Parakeet. - Commenters see strong potential use cases like medical transcription where privacy is critical. **The pushback** - Significantly slower than alternatives — ~2x realtime on M1 vs. ~60x for Parakeet-based tools like fluidaudio. - Apple Silicon only, limiting its audience compared to cross-platform alternatives like Handy. - Less mature than Handy (no cross-platform support, fewer model choices) and less polished than paid tools like Wispr Flow. - MacPorts users may have trouble with ffmpeg detection due to PATH issues. **By community** - hackernews (mixed): Community is interested and sees real promise in the privacy angle and model quality, but raises practical concerns about speed, maturity, and platform limitations compared to existing tools. **Hottest debate:** Whether Qwen3-ASR's quality advantages justify its much lower inference speed compared to Parakeet-based alternatives. **Open questions** - How does Qwen Scribe compare quantitatively (not just impressionistically) to Whisper, Parakeet, and other ASR models? - Would it make sense to add Qwen3-ASR support to the more mature Handy project rather than maintaining a separate app? - What is the full list of supported languages beyond the handful tested? - Will MacPorts ffmpeg support be added? **Highlights** > Handy is honestly more mature than mine — cross platform, MIT, and you can pick between Whisper and Parakeet. The real difference is just the model family. I run Qwen3-ASR through MLX instead of Whisper/Parakeet. In my own use it holds up way better with background noise and it punctuate noticeably more naturally, Take that as impression and not a measurement. Qwen3-ASR also accept vocabulary hints, which helps a lot with names and jargon. And mine does files and video with SRT export, not only dictation. Downside: Apple Silicon only. Wispr Flow is a different category — it's paid and your audio goes to their cloud. The text formatting is nicer than mine. Mine is local only and Apache-2.0, which for me is the entire point. > — [sidclaw on hackernews · 1 comments](https://news.ycombinator.com/item?id=49100131) > This is like 2x realtime on M1 vs 60x on fluidaudio with parakeet. > — [foobarqux on hackernews](https://news.ycombinator.com/item?id=49100991) > Stuff like this is what doctors should use instead of blasting patient data to a AI scribe who knows where. Im pretty sure that is illegal but it still takes place and I am pretty sure one of the reasons for the push to solve FHE...but its obviously way more practical and private to just do the processing locally. > — [smalltorch on hackernews](https://news.ycombinator.com/item?id=49101744) > way better from my point of view. I frequently have recordings made in noisy environments. Whisper does not do the job well under such conditions. Also, Qwen respects gramma a lot and splits sentences properly. > — [sidclaw on hackernews · 1 comments](https://news.ycombinator.com/item?id=49099261) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=49098260) · 39 points · 28 comments ## Similar posts on daily.dev - [MacOS: PrivateTranscriber. Private Alternative to WhisperFlow. Faster Than VoiceBox](https://daily.dev/posts/macos-privatetranscriber-private-alternative-to-whisperflow-faster-than-voicebox-2dijljypx) · Medium · 0 upvotes · 0 comments --- Tags: [#python](https://daily.dev/tags/python), [#mac](https://daily.dev/tags/mac), [#local-first](https://daily.dev/tags/local-first) [View this post on daily.dev](https://daily.dev/posts/github---vladuzh-qwen-scribe-private-local-transcription-and-system-wide-dictation-for-apple-silic-23jfjb1mk)