Show HN: BetterWispr – Free, Open-source Dictation For Mac

TL;DR

Before you orderOffer from Amazon

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

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

BetterWispr has been presented on Hacker News as a free, open-source dictation app for Macs running macOS 14 or later. It transcribes with on-device models, inserts cleaned text into the active app and offers optional meeting notes; the product page says it has no account requirement or cloud transcription fallback.

BetterWispr, a free, open-source dictation app for macOS 14 and later, has been presented in a Show HN post. Its maker says the app transcribes speech using models on the Mac, then inserts the result into the active app, with optional cleanup and app-specific writing styles.

The app is designed around a keyboard shortcut: users focus a text field, hold Option + Space while speaking and release to finish. Transcription begins after recording ends, rather than appearing word by word during speech. Users can change the shortcut or set it to toggle recording. If an app blocks automatic insertion, BetterWispr says the transcript can be copied from its dashboard instead.

The product page lists nine built-in speech models from Apple, Parakeet and Whisper, with language coverage dependent on the chosen model. It also allows users to configure an external service from Sarvam AI, Smallest AI or an OpenAI-compatible API. The app’s stated local-processing approach applies to local models; the maker says third-party connections are optional.

BetterWispr offers cleanup settings that range from keeping every word to removing English filler words, repetitions and false starts. A stronger setting can edit English for clarity using Apple Intelligence or Ollama on the Mac, according to the product page. Other features include a local dictation history, vocabulary hints, voice commands and meeting notes that distinguish the user’s microphone from audio captured from other apps.

Local Processing and User Choice

Dictation can involve sensitive personal or workplace speech, so the choice of where transcription takes place matters. BetterWispr’s maker says speech processed with its built-in models stays on the Mac and that the app has no cloud transcription fallback. Local models also work offline after they and their required assets are installed, according to the product information.

The app combines that privacy-oriented option with selectable models and editing controls. Readers can choose to preserve a transcript closely or apply cleanup, but language coverage and performance depend on the model and hardware. Apple Silicon is recommended, and the maker says model downloads take storage on the computer.

The project is released under the Apache 2.0 license, which lets users inspect the code and build it themselves. That offers a different distribution model from subscription dictation products, although the source material does not provide independent testing of accuracy, speed or comparative performance.

How the Dictation Workflow Works

The Show HN post introduces a utility for dictating directly into everyday Mac applications rather than requiring users to dictate inside a separate editor and transfer text afterward. BetterWispr says it uses accessibility permissions to place text in the focused app and microphone permission to record speech; Apple speech recognition also requires the relevant system permission.

The workflow is not live transcription: users record first, then receive text after releasing the shortcut. The app’s optional meeting notetaker records the user’s microphone as “Me” and audio from other apps as “Them,” alongside personal notes. The product page says summaries, decisions and action items can be generated on-device using Apple Intelligence or Ollama. Availability and results may depend on installed models, system support and hardware.

“Speech stays on your Mac.”

— BetterWispr product page

Accuracy, Speed and Compatibility

The supplied product information does not include independent benchmarks for recognition accuracy, latency or battery use, and it cautions that its interface illustration is not a benchmark. Actual results may vary with the selected model, language, accent and Mac hardware. The page recommends Apple Silicon but does not specify performance across individual devices.

Language coverage also differs by model. The maker says Parakeet covers English, Japanese and 25 European languages across its offerings, while Whisper has broader multilingual recognition; cleanup, styles and voice commands are described as English-only. It remains unclear from the supplied material how well the app performs across varied real-world speech or which model is best for particular users. No independent review or user-study results are provided.

The local-processing claim is specific to on-device models. Users who choose an external API provider send dictation to that service, according to the product page. The source material does not set out the providers’ separate data-retention terms. Automatic insertion may also fail in secure fields or apps that block pasting.

Installation and User Testing

Interested Mac users can install the app, grant the requested permissions and download a compatible speech model. BetterWispr says new versions are distributed through GitHub Releases using Sparkle and installed only when an EdDSA signature matches. The project invites users to inspect its source code, report issues or build the app themselves.

The next useful evidence will come from hands-on use and any updates from the project: performance across supported Macs and languages, transcription quality, and how well app-specific styles and meeting notes work in practice. The supplied announcement does not state a release date, download total or independent evaluation, so those points remain open.

Key Questions

Is BetterWispr free to use?

According to its maker, BetterWispr is free and open source under Apache 2.0, with no dictation subscription or account requirement. Optional external API connections may have separate provider terms or costs.

Which Macs can run BetterWispr?

The product page lists macOS 14 or later as the requirement and recommends Apple Silicon for local model processing. Model availability and performance depend on the Mac and selected language.

Does it send speech to the cloud?

The maker says built-in local models process speech on the Mac and have no cloud transcription fallback. If users configure an external API provider, dictation is sent to that service; its terms are separate.

Does text appear while I am speaking?

No. The product information says transcription starts after recording stops. Users hold Option + Space while speaking and release the keys to finish.

Which languages and editing features are supported?

Language support varies by model: the maker lists Parakeet options for English, Japanese and European languages, and broader multilingual recognition for Whisper. Cleanup, writing styles and voice commands are described as English-only.

Source: hn