OpenWispr

This program is an open-source desktop dictation application that converts speech into text. It uses OpenAI's Whisper model and can process audio locally or through cloud services. It offers a global hotkey to start and stop voice capture, automatic pasting at the cursor position, and a movable dictation panel. It supports various AI backends like OpenAI, Anthropic Claude, Google Gemini, and local models.

★★★★★

5.0(1 ratings)

Download OpenWispr (Official links)

File size: 115 MB

The latest version of OpenWispr is: 1.0.14

Operating system: Windows, Linux, MacOS

Languages: English

Developer: HeroTools

Price: $0.00 USD

Transcription with Whisper. The core of the program relies on OpenAI's Whisper model for speech-to-text conversion. Transcription runs directly on the user's device or is sent to configured cloud services, depending on the selection and available resources.
Global Hotkey. A user-defined key combination activates and deactivates audio recording from any application or window in the operating system. This method provides immediate control of the dictation process without needing to change window focus.
Fn/Globe Key Listener on macOS. Includes a specific component for macOS systems that monitors the pressing of the Fn or Globe key. This listener is pre-compiled during the program's build process and is integrated into the application packages for that system.
Automatic Pasting at the Cursor. The transcribed text is automatically inserted at the current text cursor position in the active application. This action simulates manual typing and maintains the user's workflow.
Movable Dictation Panel. A floating window displays the dictation status and the transcribed text in real-time. The user can reposition this panel on the screen according to their display preferences.
Support for Multiple AI Backends. The architecture allows for configuring and using different services for transcription and text processing. Options include cloud services from OpenAI, Anthropic Claude, and Google Gemini.
Support for Local Reasoning Models. In addition to transcription, the program can interact with locally-run language models. Supported models include Qwen, LLaMA, Mistral, and OpenAI variants, for post-text processing tasks.
Configuration and API Key Management. A centralized control panel allows adjusting all program options. From this panel, API keys for cloud services and operating preferences are managed.
Local Storage of Transcriptions. An SQLite database file saves the history of all dictation sessions performed. The records are kept in the device's internal storage.
Download and Management of Local Whisper Models. The program can download the necessary versions of the Whisper model for offline operation. It includes a utility to delete these models and free up hard disk space.
Error Handling in Key Listener. Implements a robust system for initializing the global key monitor. In case of failure on macOS, it shows a warning dialog informing that the Globe key shortcut is disabled, and prevents duplicate error reports.
Backend Integration During Compilation. The program's build process pre-compiles and packages the macOS-specific key listener component. The resulting binary is included within the application bundle distributed to the end user.

The development of OpenWispr began following the public release of OpenAI's Whisper model. The goal was to create a practical desktop interface that leveraged this technology. The program is written primarily in JavaScript/TypeScript using the Electron framework, allowing it to run on Windows, macOS, and Linux. The first public version was released in the year 2023. The development is carried out by a small team of independent programmers who focus on AI-based productivity tools.

Alternatives to OpenWispr:

Claritykey — Free Download. AI writing assistance for dyslexia and communication barriers