# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/story/show-hn-betterwispr-free-open-source-dictation-for-mac-50042333

# Show HN: BetterWispr – Free, open\-source dictation for Mac

81 points · 48 comments

[Full discussion](<https://news.ycombinator.com/item?id=50042333>)

[Read original](<https://betterwispr.com/>)

Category: [Models & Products](<https://hacksnap.live/?category=models-products>)

## Skept-o-meter & Hotness

Skept\-o\-meter: Low\. Estimated from 5 comments\.

5 comments for the summary\.

Peak rank: \#5

Time in Top 10: 3\.0 hours

Hacksnap ranks recent stories first, then orders each group by points\. Peak rank uses all retained observations\. Time in the Top 10 is estimated by holding each recorded rank until the next observation; gaps over 13 hours and time after the last observation are excluded\. Movement between observations is unknown\.

4 recorded rank observations from 2026\-10\-11T18:00:46\.396018\+00:00 to 2026\-10\-11T21:00:51\.174035\+00:00\.

Hotness — latest 4 recorded Hacksnap ranks:

2026\-10\-11T18:00:46\.396018\+00:00: rank \#5

2026\-10\-11T19:01:19\.364382\+00:00: rank \#6

2026\-10\-11T20:00:59\.17002\+00:00: rank \#6

2026\-10\-11T21:00:51\.174035\+00:00: rank \#5

BetterWispr’s local\-first Mac dictation draws interest for voice\-first coding, but the small sample questions how it differs from Handy and VoiceInk and asks for tighter agent/IDE integration\.

## The brief

BetterWispr is a free, Apache\-2\.0 macOS dictation app that runs speech recognition locally and inserts cleaned text into any app\. It uses a hold\-Option\-Space shortcut, offers nine built\-in models including Apple speech, Parakeet and Whisper, and adds per\-app writing styles, voice commands, vocabulary learning and meeting notes\. The page claims no account or cloud fallback, with optional local LLM cleanup via Apple Intelligence or Ollama and optional external API providers if selected\.

- Hold Option\+Space in any app to record; release to transcribe, or switch to press\-to\-toggle or a custom shortcut\. Transcription begins after recording stops, not word\-by\-word\.
- Nine built\-in models include Apple speech, Parakeet TDT v3 (25 European languages), Parakeet Japanese and Whisper Large v3 Turbo/WhisperKit; users can also plug in Sarvam AI, Smallest AI or OpenAI\-compatible APIs\.
- Cleanup levels range from Light (removes English fillers, repeats, false starts) to Medium (local LLM clarity edits) and None; per\-app styles adjust punctuation and capitalization, and voice commands handle punctuation, new paragraphs and 'scratch that'\.
- Notetaker transcribes the mic as Me and other app audio as Them on macOS 14\.2\+, then Apple Intelligence or Ollama can summarize decisions and action items locally; Claude Code, Codex and OpenAI\-compatible endpoints are optional and send transcripts to those providers\.
- Privacy claims include local models, no account, no cloud fallback, on\-Mac history with raw/corrected text, and Sparkle updates from GitHub Releases verified by EdDSA signature\. Requirements are macOS 14\+, Apple Silicon recommended, with microphone, accessibility and speech\-recognition permissions\.
- Automatic paste may be blocked in secure fields or some apps; BetterWispr then copies the transcript to the clipboard\. Filler cleanup, styles and voice commands work in English\.

## Discussion themes

Analyzed: 2026\-10\-11T19:01:17\.43934\+00:00

Analysis sample: Based on 5 of 6 usable stored comments. Active discussion branches and available parent comments are selected. The analysis input was further shortened to fit its context limit.

This sample may omit parts of the full thread. Selected themes do not measure community opinion or how common a view is.

### Overlap with Handy and VoiceInk

Zizizizz says BetterWispr resembles Handy in features and asks for its use case over that project; crazygringo cites VoiceInk as a longer\-standing similar tool\. No direct differentiator is established in the sample\.

Sources: [Comment 50043741](<https://news.ycombinator.com/item?id=50043741>) · [Comment 50045019](<https://news.ycombinator.com/item?id=50045019>)

### Whisper remains the default STT model

Royce\-CMR says many dictation competitors still use OpenAI Whisper, released in 2022, and finds it surprising that it remains the default all\-rounder in 2026; sunnybeetroot asks whether anything better exists\. No alternative model is named in the replies\.

Sources: [Comment 50045439](<https://news.ycombinator.com/item?id=50045439>) · [Comment 50045882](<https://news.ycombinator.com/item?id=50045882>)

### Agent IDE auto\-send

frenchie4111 wants dictation integrated into an agent IDE and needs an end\-of\-message action, such as a newline, to auto\-send chat; suggests a per\-active\-app setting\.

Sources: [Comment 50044781](<https://news.ycombinator.com/item?id=50044781>)

### Remote\-control and command\-layer workflow

crazygringo remaps a Bluetooth media remote via Karabiner Elements for dictation, deletion, newline, submit and scrolling, and wants a second button for local LLM commands instead of macOS Voice Control\.

Sources: [Comment 50045019](<https://news.ycombinator.com/item?id=50045019>)

## Sources & coverage

AI-generated summary · 2026\-10\-11T19:01:17\.909983\+00:00

Based on 5 of 6 usable stored comments, selected by depth and branch activity. This is a sample of the discussion. The model input was further shortened to fit its context limit. Article text may also be shortened.

Generated using deepseek\-ai/DeepSeek\-V4\.1\-Flash. Check the linked sources for full context.
