TL;DR
Get the latest gadgets 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. Its developer says speech can be transcribed on the Mac without an account or cloud fallback, with model, cleanup and privacy settings controlled by the user.
BetterWispr, a free, open-source dictation app for Mac, has been presented in a Show HN launch, offering users a keyboard shortcut to speak into text fields and transcribe speech on their computer. The developer says the app supports macOS 14 or later, requires no account and has no cloud transcription fallback when using its local models.
The app is designed to work across applications: users focus a text field, hold Option + Space while speaking, then release the shortcut to insert the transcript. The developer says users can change the shortcut or configure it to toggle recording. Transcription starts after recording stops, rather than appearing word by word as the user speaks.
BetterWispr includes nine built-in speech models from Apple Speech, Parakeet and Whisper, according to its website. Language coverage depends on the selected model. The app also offers the option to connect to Sarvam AI, Smallest AI or an OpenAI-compatible API; those are separate choices from local processing and may involve sending speech to an outside provider.
Users can choose how much the app edits a transcript. The default Light setting removes English filler words, repetitions and false starts, while None retains the spoken words. Medium can edit English for clarity using a selected notes model, such as Apple Intelligence or Ollama running on the Mac. Users can also set writing tones for different apps, add names and terms, and correct recognized words through the app’s history.
Local Dictation Without a Subscription
BetterWispr is aimed at people who want speech-to-text across everyday Mac apps without a dictation subscription or an account. Its Apache 2.0 license allows users to inspect the source code, report issues or build the app themselves, according to the developer.
The local-processing option may appeal to people who prefer to keep recordings on their device. The developer says installed local models can work offline and that speech does not need to be sent to a transcription server. That claim applies to the local workflow: the app also lists optional third-party API connections, and the user’s choice of model and settings affects where processing occurs.
Practical access depends on hardware and language. The site recommends Apple Silicon for local inference and says model performance varies by Mac and selected language. Automatic insertion also requires permissions, and the developer says some apps or secure fields may block pasting.
As an affiliate, we earn on qualifying purchases.
How the Mac Workflow Works
The launch describes a three-step workflow: install the app and a speech model, place the cursor in a text field, then hold a shortcut to record and release it to transcribe. Users grant microphone permission to record; accessibility permission supports insertion into other apps, while Apple Speech also requires speech-recognition permission.
The product includes a meeting notetaker that records the user’s microphone as “Me” and audio from other apps as “Them,” alongside personal notes. The developer says Apple Intelligence or Ollama can generate summaries, decisions and action items locally. BetterWispr also stores raw and corrected dictations in local history, which users can turn off. Updates are distributed through GitHub Releases using Sparkle, with the developer describing signature checks for installations.
“Speech stays on your Mac.”
— BetterWispr website
As an affiliate, we earn on qualifying purchases.
Performance and Privacy Details
The supplied launch material does not give independent benchmark results for transcription accuracy, processing speed, battery use or model performance on particular Mac configurations. A demonstration on the product site is explicitly described as an illustration, not a benchmark. The developer recommends Apple Silicon but does not provide comparative hardware measurements.
The source also does not provide a dated release record, user numbers or independent security and privacy testing. Local model use is described as offline after installation, but optional API connections can send dictation to the selected provider. The exact data handling of those external services is not detailed in the supplied material.
Language support and editing features have limits: model coverage varies, and the developer says cleanup, styles and voice commands work in English. It is also unclear from the launch material how well the app handles accents, noisy rooms or specialized vocabulary across different models.
As an affiliate, we earn on qualifying purchases.
Installation and User Testing
Readers can review the code and download BetterWispr through the project’s GitHub links, according to the developer. The immediate next step for prospective users is to check their Mac’s operating system and hardware, install a compatible model, and test it with their own language and speech before relying on it for regular work.
Further evidence about reliability and adoption will depend on user testing and any later product updates. The developer invites people to report issues, but the source material does not announce a release timetable or a specific upcoming milestone.
As an affiliate, we earn on qualifying purchases.
Key Questions
Is BetterWispr free?
Yes. The developer says it is free and open source under the Apache 2.0 license, with no dictation subscription, account or API key required for the local option. Downloaded model files use storage on the Mac.
Which Macs can run BetterWispr?
The stated minimum is macOS 14. Apple Silicon is recommended for local model processing, while model availability and performance depend on the hardware and language.
Does BetterWispr send my speech to the cloud?
The developer says its built-in local models transcribe on the Mac and have no cloud transcription fallback. Users can also choose third-party API connections, which are distinct from local processing and may send dictation to the selected provider.
Does the transcript appear as I speak?
No. According to the product website, recording ends when the user releases the shortcut, and transcription begins afterward, rather than appearing word by word during speech.
What happens if BetterWispr cannot paste into an app?
The developer says some apps and secure fields may block automatic insertion. In that case, users can retrieve the transcript from the app and copy it manually; the app also says it copies text to the clipboard when insertion is blocked.
Source: hn
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
