TL;DR
Get everyday helpers delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
BetterWispr has introduced a free, open-source dictation app for Macs running macOS 14 or later. It transcribes speech using models installed on the Mac, inserts the result into the active app and offers optional cleanup, meeting notes and configurable shortcuts.
BetterWispr has launched as a free, open-source dictation app for Macs running macOS 14 or later, offering local speech recognition and text insertion into the app a user is working in. The product’s website says it requires no account and can work offline after compatible models are installed, giving Mac users an alternative to dictation tools that rely on cloud transcription.
The app is designed around a keyboard shortcut: users place the cursor in a text field, hold Option + Space while speaking, then release to transcribe. The website says the text appears after recording stops, rather than being transcribed word by word in real time. Users can switch to a press-to-toggle mode or configure another shortcut.
BetterWispr includes nine built-in speech models from Apple Speech, Parakeet and Whisper, according to the product site. It also lets users connect their own Sarvam AI, Smallest AI or OpenAI-compatible API. Those external services are distinct from the default local setup: the site says local models run on the Mac and there is no cloud transcription fallback. Downloading model files requires storage, and a network connection is needed to install them.
After transcription, users can choose how much text cleanup to apply. The default Light setting removes English filler words, repetitions and false starts; None preserves the words as spoken, while Medium also edits for clarity using a selected notes model. The site says cleanup, writing styles and voice commands are limited to English. Per-app styles can adjust tone, and spoken commands can add punctuation, start a new paragraph or delete the last sentence.
Local Dictation Without a Subscription
The launch gives Mac users a no-cost, open-source option for converting speech into text across apps, with the main speech-processing path running on the computer. Keeping recordings and transcripts local may appeal to people who do not want routine dictation sent to a transcription service, though the product’s broader integrations mean users should check which processing option they select.
The project is licensed under Apache 2.0, and its site says people can inspect the code, report issues or build the app themselves. That makes the implementation more open to review than a closed-source tool, but an open license by itself does not establish how well transcription performs or how the app behaves on every Mac. The source provides no independent accuracy tests or comparative results.
BetterWispr also combines dictation with meeting transcription and notes. Its Notetaker is described as separating microphone audio as “Me” from other app audio as “Them,” with summaries, decisions and action items generated using Apple Intelligence or Ollama on the Mac. These features extend the app beyond short voice input, but their usefulness will depend on the selected models, hardware and recording conditions.
As an affiliate, we earn on qualifying purchases.
How BetterWispr Handles Speech
BetterWispr’s workflow has three stages: install the app and a compatible local speech model, focus a text field, then hold a shortcut and speak. The website says the model downloads once and can subsequently work offline. Apple Silicon is recommended for local model inference, while language availability and performance depend on the chosen model and the Mac.
The app asks for permissions that support its functions. Its site lists microphone access for recording, Accessibility permission for inserting text into other apps and Speech Recognition permission when using Apple Speech. Automatic insertion may not work in secure fields or apps that block pasting; the site says users can retrieve the transcript and copy it manually if insertion fails.
Dictation history can store both raw and corrected text locally, and users can disable history. BetterWispr also says it can learn names and terms from corrections. Its Insights panel counts dictated words, pace, cleanup and app use from local history. The site describes the typing-versus-speaking graphic as an illustration, not a benchmark, and gives simulated rates of 40 words per minute typing and 150 words per minute speaking.
“Speech stays on your Mac.”
— BetterWispr website
As an affiliate, we earn on qualifying purchases.
Accuracy and Compatibility Still Vary
The available source is the product’s own website and launch material; it does not include independent testing of transcription accuracy, latency, battery use or performance across different Mac models. The site says support and results vary by hardware, language and model, so the practical experience cannot be established from the launch description alone.
The product describes its local mode as having no cloud transcription fallback, but users can choose external APIs for some functions. The source says those connections send dictation to the selected provider, but does not provide a full technical account of every data flow or explain all provider-specific retention practices. The exact publication date of the Show HN post is also not supplied.
As an affiliate, we earn on qualifying purchases.
Installation and User Testing
Interested Mac users can install BetterWispr from the project’s release channel, follow its setup guide, grant the requested permissions and download a compatible speech model. The project says releases are distributed through GitHub Releases using Sparkle and installed only when an EdDSA signature matches.
The next useful evidence will come from hands-on testing and project updates: whether the app works reliably across target applications, how its models perform for different accents and languages, and how much local resources they use. The source does not announce a test schedule or future release date, so those developments remain open.
As an affiliate, we earn on qualifying purchases.
Key Questions
Is BetterWispr free to use?
The product says BetterWispr is free and open source under Apache 2.0, with no dictation subscription, account or API key required for its built-in local workflow. Model downloads use storage on the Mac.
Does BetterWispr send speech to the cloud?
According to the product site, built-in local models process speech on the Mac, with no cloud transcription fallback. Users can choose external API connections for some functions; the site says those send dictation to the selected provider.
Which Macs and languages does it support?
BetterWispr requires macOS 14 or later, with Apple Silicon recommended. Language support depends on the model: the site lists Parakeet options for Japanese and 25 European languages, and says Whisper supports a broader range. Apple Speech depends on locale and installed assets.
Does it transcribe while I am speaking?
No. The product says transcription starts after recording stops. The default method is to hold Option + Space while speaking and release the keys to finish.
Can I keep every word exactly as spoken?
Yes. The site says users can select None for cleanup to preserve their words, or choose Light or Medium cleanup. Light removes English fillers, repeats and false starts; Medium also edits for clarity using a selected notes model.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
