Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Voice, Speech & Multimodal

OpenWhispr Review: Open-Source Dictation and Meeting Notes App

·7 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

OpenWhispr (version 1.10.4) is an MIT-licensed voice-to-text app that pastes your dictation into any app, and whether your voice stays on your device depends on whether you pick a local or a cloud speech model. It adds meeting transcription, notes and an AI assistant, and every statement here comes from the maker's website, privacy policy, documentation and GitHub repository.

Key Takeaways

  • Maintenance: the main branch last changed on 23 October 2026 and version 1.10.4 was released on 22 October 2026, so no stale-project warning applies.
  • Cost: local dictation is free and unlimited; cloud and team features sit behind paid plans.
  • Engines: whisper.cpp runs the Whisper models and sherpa-onnx runs the other local models.
  • Platforms: desktop installers for macOS, Windows and Linux, plus a free iPhone and iPad app.
  • Operator: Gizmo Labs Inc., a Delaware company; the code is open under the MIT license.

📍 In One Sentence

OpenWhispr is an open-source desktop and mobile app that turns speech into text with on-device or cloud models, pastes it into the active app and can record and label meetings.

💬 In Plain Terms

Think of it as a voice keyboard for your whole computer, with a notebook that can also listen to your meetings.

📌Note: This review is based on openwhispr.com, its privacy policy, the documentation and the GitHub repository, checked on 24 October 2026. PromptQuorum has not installed or timed the app, so accuracy and speed are not assessed.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

OpenWhispr is an open-source voice-to-text app for macOS, Windows, Linux, iPhone and iPad: you press a hotkey, speak, and the text is pasted where your cursor is, with optional meeting transcription and notes. You choose between speech models that run on your own device and cloud processing, and that choice decides where your voice goes. This review is based on the website, its privacy policy, the documentation and the GitHub repository, checked on 24 October 2026; PromptQuorum has not tested the app hands-on.

OpenWhispr Review: Open-Source Dictation and Meeting Notes App

What Is OpenWhispr?

OpenWhispr is a voice-to-text app that records while you press a hotkey, transcribes the audio and pastes the text into the focused app. The README calls it "the open-source and free alternative to WisprFlow and Granola", and the website adds meeting notes, an AI assistant and 100+ languages. The repository was created in June 2025.

The app is published by Gizmo Labs Inc., a Delaware corporation, which also runs the paid OpenWhispr Cloud. The desktop app is built with Electron; the iPhone and iPad app is a separate Expo project in the same repository.

Get It

OpenWhispr is a download for macOS, Windows and Linux, plus an iPhone and iPad app on the App Store; the documentation says there is no web version and no browser extension.

Option
How to get it
macOS.dmg (Apple silicon, Intel): openwhispr.com/download
Windows.exe installer (x64): openwhispr.com/download
Linux.AppImage, .deb, .rpm, .tar.gz (x86_64): openwhispr.com/download
iPhone and iPadApp Store, iOS 17.0 or later
Websiteopenwhispr.com
Privacy policyopenwhispr.com/privacy
Source codegithub.com/OpenWhispr/openwhispr, MIT license

Version reviewed: 1.10.4, published on 22 October 2026 per the GitHub Releases page, which lists every desktop installer; the iPhone and iPad app is version 1.2.4 in its App Store listing and showed as free in the US, German, French, Spanish, Brazilian, Japanese and Korean storefronts. A community-maintained Homebrew cask (brew install --cask openwhispr) also exists. Paid plans are covered under Plans.

Requirements

The documentation lists macOS 12, Windows 10 and a 64-bit Linux as the minimum; running speech models on your own device is the demanding part.

Operating system

What the sources say:
macOS 12 or later (Apple silicon or Intel), Windows 10 or later (64-bit x64), Linux 64-bit x86-64 with PulseAudio or PipeWire; no ARM builds for Windows or Linux

Speech models

What the sources say:
Downloaded in the app: Orukeet about 672 MB (the recommended default), Whisper from about 75 MB (tiny) to about 3 GB (large), Parakeet about 631 to 680 MB, Cohere Transcribe about 1.7 GB

Hardware

What the sources say:
Local Whisper can use Metal on Apple silicon, CUDA on NVIDIA GPUs or Vulkan on AMD and Intel GPUs, and falls back to the CPU if the GPU runtime fails; no RAM figure is given

Permissions

What the sources say:
Microphone and accessibility access for pasting text; macOS 14.2 or later plus the System Audio Recording permission to capture the other side of a meeting

Intel Macs

What the sources say:
Live speaker identification and voice fingerprinting are unavailable; meetings still record and transcribe, and note search falls back to keywords (README)

Internet

What the sources say:
Needed to download models and for cloud features; local transcription works offline once the model is on disk

Model sizes are the figures in the documentation's local-models guide; the in-app list may differ.

Features Confirmed by the Sources

Every item below comes from the website, the documentation or the README; none has been independently tested.

  • Dictation anywhere. A global hotkey records, transcribes and pastes into the focused app; a dedicated hotkey dictates in one language and pastes in another.
  • Two local engines. Whisper runs through whisper.cpp, while the Orukeet, Parakeet and Nemotron families run through sherpa-onnx; the streaming Nemotron models can show partial text as you speak.
  • Meetings. The app detects Zoom, Teams and FaceTime calls, labels speakers on the device and can link Google, Microsoft or Apple Calendar events to notes.
  • Notes and import. Notes have folders and a local embedding model for semantic search; audio and video files, or a pasted YouTube or audio URL, can be transcribed.
  • Assistant and integrations. A voice-assistant hotkey sends what you say to a model you pick, cloud or local; a public API and an MCP server connect other tools.

Plans

Local dictation is free and unlimited on every plan; the paid plans add cloud transcription, more meeting hours and team features. All figures are the US-dollar prices on the pricing page.

Free

What the pricing page lists:
$0: unlimited local models, unlimited cloud transcription with your own API keys, 2,000 words per week of OpenWhispr Cloud transcription, 5 hours of meeting recordings per month

Pro

What the pricing page lists:
$80 per user per year billed annually ($6.67 per month shown): unlimited cloud transcription, agent mode, 20 hours of meetings per month, sync, API and MCP access, iPhone and iPad app

Business

What the pricing page lists:
$160 per user per year ($13.33 per month shown): everything in Pro plus unlimited meeting recordings, chat over your data and priority support

Enterprise

What the pricing page lists:
Custom: SSO, SAML and SCIM, audit logs, retention controls, dedicated support

Figures as shown on 24 October 2026; no other currency or regional price was checked, and taxes are not included.

Data Flow and Privacy

The privacy policy, last updated 12 October 2026, describes three transcription modes on the desktop: local processing, where no audio is transmitted anywhere; your own API key, where audio goes directly to the provider you chose; and OpenWhispr Cloud, where audio is sent to its API or a speech-to-text provider, processed in real time and not stored by OpenWhispr. Cloud usage records (word count, provider, model, length) are kept for billing, and the policy says your data is not used to train models. Transcribed text lives in a local SQLite database, and the desktop app keeps dictation audio on your device for 30 days by default (up to 90, or off).

Cloud backup is off by default on the desktop and on by default for paid plans in the mobile app. Notes stored in OpenWhispr's cloud are sent to OpenAI for search indexing, and web searches run by the assistant go through the search provider Exa. The README says "no telemetry", while the policy lists analytics, attribution and purchase SDKs in the mobile app (AppsFlyer, RevenueCat, Superwall), with usage analytics switchable off.

📌Note: These are the maker's declarations, not audit results. The code is public under the MIT license, but PromptQuorum has not read it or inspected the app's network traffic.

What We Could Not Verify

  • Accuracy and speed. PromptQuorum did not measure word error rates, latency or language coverage on any model.
  • Compliance badges. The website shows HIPAA, ISO 27001, SOC 2 Type II and GDPR badges and links a Trust Center; we did not read any audit report or certificate.
  • Download figure. The website says "Downloaded by 150k+ users"; we have no way to check that.
  • Which iPhone features need a plan. The App Store listing offers a free download, while the pricing page lists the mobile app under the paid plans.
  • Installer signing. Whether the macOS and Windows installers are notarized or code-signed was not checked.

Competitors and Alternatives

Handy

Platforms:
macOS, Windows, Linux
Price / license:
Free / MIT
Key difference:
Narrower free dictation tool: no meeting notes, no paid cloud plan
Articles about Handy (2)

Also mentioned in:

+5 more not shown

MacWhisper

Platforms:
macOS
Price / license:
Freemium / Proprietary
Key difference:
Mac-only app with a one-time purchase for local Whisper and Parakeet transcription; closed source
Articles about MacWhisper (2)

Also mentioned in:

AquaVoice

Platforms:
macOS, Windows, iOS, Android
Price / license:
Freemium / Proprietary
Key difference:
Cloud dictation service that types into the active app, so it is not an offline option
Articles about AquaVoice (2)

These are same-category apps from the PromptQuorum directory; confirm each app's current price, license and platforms on its own page.

Compare in directory

Frequently Asked Questions

Which local model should I start with?

The documentation recommends Orukeet, a multilingual Parakeet fine-tune from Oruk, and a fresh install lands on it when you choose the local route; for Whisper it recommends the base size. It calls Cohere Transcribe its most accurate offline model, at a 5.42% average word error rate across 14 languages, a project figure we did not check.

Does it support languages other than English?

The website claims 100+ languages with auto-detection, but local coverage depends on the model: the documentation lists 25 languages for Parakeet v3, 15 for one Nemotron model and English only for two others. We did not verify the 100+ figure.

Can I use my own API keys?

Yes on the desktop: the Free plan lists unlimited cloud transcription with your own API keys, and the privacy policy says that audio then goes directly to your provider without passing through OpenWhispr. The iPhone listing also mentions connecting your own provider.

Is there a command line or a developer API?

The pricing page lists a command-line interface that works locally on the Free plan and locally and remotely on paid plans, and API and MCP access from Pro upward; the documentation links an API overview.

How often is it updated?

Version 1.10.3 was published on 12 October 2026 and 1.10.4 on 22 October 2026, and the repository keeps a changelog. Releases that frequent mean settings and plan limits can change, so check the pricing page before relying on a figure here.

Verdict

OpenWhispr is an actively updated, MIT-licensed way to dictate into any app and record meetings on macOS, Windows, Linux and iPhone, with speech models you can run on your own device or swap for a cloud provider. Its privacy story is only as strong as the mode you pick, the paid plans and the mobile app add cloud services and analytics that the README's "no telemetry" line does not describe, and none of its accuracy, speed or compliance badges could be checked here. It suits people who want one app for dictation and meeting notes and will keep to local models if privacy matters; anyone who needs proven accuracy or certified compliance should test it and request the audit reports first.

Sources

← Back to Power Local LLM

MCP