OpenLess: The Open-Source AI Voice Input Tool for Developers
Stop typing, start talking. OpenLess is a cross-platform, privacy-focused tool that turns your voice into structured, AI-polished text directly at your cursor.
For developers and power users, the friction between "having an idea" and "getting it into the editor" is real. While commercial tools like Wispr Flow and Typeless have popularized the "push-to-talk, get-polished-text" workflow, they often come with subscription fees, closed-source black boxes, and concerns about where your data goes.
Enter OpenLess: a fully open-source, local-first alternative for macOS and Windows that brings AI-powered voice dictation directly to your cursor.
Why OpenLess Matters
OpenLess isn't just a speech-to-text tool; it’s a productivity bridge. Its primary value proposition is its AI-Prompt Mode. Instead of just transcribing your rambling, it uses an LLM to restructure your speech into a clean, context-rich prompt.
The Workflow:
- Hold your global hotkey.
- Speak your request (e.g., "I need a SQL query for last month's orders, grouped by customer, sorted by amount descending.").
- Release the key.
- Result: The app transcribes, polishes, and instantly pastes the structured prompt into your IDE, ChatGPT, or Notion.
Key Technical Features
Built with Tauri 2, Rust, and React/TypeScript, OpenLess is designed for performance and extensibility:
- Bring Your Own Model: Unlike SaaS competitors, you control the pipeline. You can configure your own API keys for Volcengine (ASR) and any OpenAI-compatible LLM endpoint (DeepSeek, Ark, Anthropic, etc.).
- Privacy-First: Your data stays on your machine. Credentials are stored in the OS-native vault (macOS Keychain, Windows Credential Manager), not in plaintext files.
- Local-First Architecture: The app supports local ASR via bundled Qwen3-ASR models, ensuring you can dictate even without a high-latency cloud connection.
- Smart Dictionary: You can define custom terms, product names, or technical jargon. These are injected into the ASR context as hotwords, significantly improving accuracy for developer-specific vocabulary.
- Clipboard Fallback: If the app detects that it cannot inject text into the current window (due to security restrictions or app-specific limitations), it automatically copies the result to your clipboard, ensuring you never lose your work.
Getting Started
OpenLess is designed for developers who want to tinker and users who want a stable tool.
For End Users
You can grab the latest release from their GitHub Releases.
- macOS:
brew install --cask openless - Windows: Use the provided
.exeinstaller.
Note: After installation, ensure you grant the necessary Accessibility and Microphone permissions in your OS settings.
For Developers
If you want to contribute or build from source, the project uses a clean modular structure. To get started:
# Clone and initialize submodules
git submodule update --init --recursive
# Install dependencies
cd openless-all/app
npm ci
# Run in development mode
npm run tauri dev
The "Structured" Difference
Most voice-to-text tools aim for verbatim transcription. OpenLess aims for utility. By using a "Polish" provider, it strips away filler words ("um," "uh," "so"), fixes punctuation, and organizes your thoughts into bullet points or code blocks. It doesn't try to answer your questions—it simply prepares your input so that the actual AI (ChatGPT/Claude) can provide the best possible response.
If you are tired of the subscription fatigue of modern AI tools and want a transparent, hackable, and fast voice-input solution, OpenLess is a project worth watching.