TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
BetterWispr has introduced a free, open-source dictation app for macOS 14 and later. It records speech through a keyboard shortcut, transcribes it using a selected model, and can clean up and insert text locally; the developer says speech is not sent to a cloud transcription service by default.
BetterWispr has launched as a free, open-source dictation app for Mac, letting users speak into a keyboard shortcut and place a transcribed, optionally cleaned-up version of their speech in the app they are using. The developer says speech processing runs on the Mac with built-in models, without an account or cloud transcription fallback; the app requires macOS 14 or later, with Apple Silicon recommended.
Users focus a text field, hold Option + Space while speaking, then release the keys to finish recording. BetterWispr transcribes after recording stops rather than displaying words as the person speaks. Users can change the shortcut, including setting it to toggle recording, and can choose among nine built-in models from Apple speech, Parakeet and Whisper. The developer also lists options to connect to Sarvam AI, Smallest AI or an OpenAI-compatible API.
Text cleanup is adjustable. The default Light setting removes English filler words, repetitions and false starts; Medium also edits for clarity using a selected notes model, while None keeps the words as spoken. Users can set different writing styles for different apps, such as casual or formal, and add names and terms as recognition hints or replacements. The company says styles, voice commands and filler cleanup work in English.
The app also includes meeting notes that can transcribe the user’s microphone and audio from other apps separately, labeling them “Me” and “Them,” alongside the user’s notes. Apple Intelligence or Ollama can produce summaries, decisions and action items on the Mac, according to the product site. Dictation history can be disabled, and the developer says it stores original and corrected text locally. BetterWispr is released under the Apache 2.0 license; its site says updates are distributed through GitHub Releases and checked against an EdDSA signature.
Local Dictation With User Controls
BetterWispr is aimed at people who want speech input without routinely sending recordings to a transcription service. Its local-processing approach may appeal to users handling sensitive drafts or meeting notes, though privacy depends on which models and optional services they choose. The developer says built-in speech models run on the Mac and work offline after installation; models must first be downloaded, and some optional integrations can send dictation to an outside provider if selected.
The app also combines transcription with editing controls that are often treated as separate tasks: users can choose how much speech is cleaned up, set a tone by app and correct recognized terms. That offers more control over whether dictation preserves a speaker’s wording or produces a more polished draft. These are product features, not independent evidence of accuracy or performance. The site gives no comparative recognition benchmark, and its typing-speed illustration is explicitly labeled as illustrative rather than a benchmark.
For users, the practical trade-offs include model downloads, local storage and hardware-dependent performance. Apple Silicon is recommended, and the developer says model availability and speed depend on the computer and language. Users also need to grant microphone and Accessibility permissions for recording and automatic text insertion; some secure fields or apps may block insertion.
As an affiliate, we earn on qualifying purchases.
How BetterWispr Handles Speech
BetterWispr is presented through a Show HN launch and the product’s own website. Its central workflow is push-to-talk: record a phrase, stop recording, transcribe it, then insert the result in the focused app. This differs from live dictation that displays words as the user speaks. The developer says that if an app blocks automatic paste, the transcript can be retrieved from the dashboard and copied manually.
Language support varies by model rather than being uniform across the app. The product site says Parakeet covers English, Japanese and 25 European languages across its offerings, while Whisper supports a broader range of languages; Apple speech depends on the user’s locale and installed system assets. The app’s English-only cleanup and voice-command functions are separate from the speech recognition models, so support for a language does not necessarily mean every writing feature is available in it.
The website describes BetterWispr as free, with no required account, dictation subscription or API key for its built-in local features. Optional external services are available, but the source does not provide detailed technical documentation in the launch material about their data handling. Users can inspect the source code under the Apache 2.0 license, according to the developer.
“Speech stays on your Mac.”
— BetterWispr website
open source speech recognition software Mac
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Performance and Privacy Details
The launch material does not report independent testing of transcription accuracy, speed, battery use or performance across Mac models. It also does not provide a direct comparison with other dictation products. The site’s illustrative typing graphic should not be treated as evidence of a measured time saving.
It remains unclear how the app performs for different accents, noisy environments and less common languages. The developer says results depend on the selected model, language and hardware, and advises users to test a model with their own speech. Specific model sizes, storage requirements and comparative resource use are not stated in the supplied launch information.
The local-processing claim applies to the built-in workflow as described by the developer. The site also allows users to choose external APIs, and says some editing connections send dictation to the selected provider. The source material does not spell out the technical data flows for every optional service or model. Users should check the settings and provider terms before enabling an external connection.
As an affiliate, we earn on qualifying purchases.
Install, Test and Track Updates
Interested Mac users can download BetterWispr, complete its permission setup and install a local model before testing it in a text field. The app requires macOS 14 or later; Apple Silicon is recommended. Users can compare the None, Light and Medium cleanup settings, check model support for their language, and confirm whether text insertion works in the apps they use. BetterWispr says speech recognition can work offline after the compatible model and its required assets have been installed.
The developer says releases are distributed through GitHub and installed only when their EdDSA signature matches. The source material does not announce a future release date, roadmap or independent evaluation. Any assessment of accuracy, reliability and time savings will depend on testing beyond the launch description.
As an affiliate, we earn on qualifying purchases.
Key Questions
Is BetterWispr free?
Yes. The developer says the app is free and open source under the Apache 2.0 license, with no required account, dictation subscription or API key for the built-in features. Downloaded model files use storage on the Mac.
Does BetterWispr send recordings to the cloud?
The developer says built-in speech models process dictation on the Mac and that there is no cloud transcription fallback. Once a compatible model is installed, local transcription can work offline. Optional API connections may send dictation to the selected provider, according to the product site.
Which Macs and languages does it support?
BetterWispr requires macOS 14 or later, and Apple Silicon is recommended. Language support depends on the model: the site lists Parakeet options for English, Japanese and European languages, broader multilingual recognition from Whisper, and Apple speech support based on locale and installed assets.
Does it show a transcript while I speak?
No. BetterWispr records while the shortcut is held, then transcribes after recording ends. The product site says words do not appear in real time while the user is speaking.
Can users control how much the app edits their words?
Yes. Users can choose None to preserve every word, Light to remove English filler words, repetitions and false starts, or Medium to also edit for clarity. The app also offers per-app writing styles, but the site says cleanup and styles apply to English.
Source: hn
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
