OpenLess

Category: AI Voice Input & Dictation App (ASR + LLM polishing)

What it is

OpenLess is a cross-platform voice-input application (macOS, Windows, Linux, Android) that works like a smarter, open-source alternative to tools like Wispr Flow or Superwhisper. You press a global hotkey, speak, and the app records your audio, converts it to text, uses an AI model to clean it up or structure it, and then inserts the polished text directly into the cursor's location (e.g., in ChatGPT, Notion, an email, or a code editor).

Its headline feature is AI-prompt mode: speak freely and messily, and OpenLess will turn your loose thoughts into a structured, detailed prompt that you can copy/paste into ChatGPT, Claude, or Cursor.

How it works

• Pipeline: Speech → Transcription → AI Cleanup/Polish → Insert at cursor (falls back to clipboard if insertion is blocked).
• Bring-your-own AI: Supports a dozen-plus ASR providers (Volcengine, iFlytek, Alibaba, OpenAI Whisper, ElevenLabs, etc.) and LLM polish providers (OpenAI, DeepSeek, Google Gemini, etc.). You plug in your own API keys—no vendor lock-in.
• Local-first: The app is open source and can run fully locally (e.g., bundled Qwen3-ASR models on macOS, local Whisper on Windows), so your data doesn't have to leave your machine.
• Streaming Insertion: Text is typed to your cursor character-by-character as it is polished, reducing perceived latency.
• Hotkey integration: Uses native OS hooks (CGEventTap on macOS, low-level keyboard hook on Windows) for a reliable global hotkey.

Key features

• AI-prompt mode: Structures rambling speech into a prompt with context, constraints, and a clear request.
• Style Pack Marketplace: Create or install custom writing styles (e.g., "terse engineering commit messages" or "warm customer-support replies") with custom system prompts, switchable by hotkey.
• Learning Dictionary: When you correct a word it misheard, it asks if you want to remember it. Suggestions are always reviewed by you, never learned silently.
• Selection-ask QA panel: A separate hotkey opens a panel to run voice Q&A against highlighted text in any app.
• Translation hotkey: Converts speech directly into a configured target language.
• Persistent tray icon & floating mini status capsule: Follows your display if you use multiple monitors.
• Multilingual UI: Supports English, Chinese, Japanese, Korean, etc.

Technology stack

• Core: Rust shared backend (openless-core).
• Desktop host: Tauri 2 with a React/TypeScript frontend (macOS, Windows, Android).
• Linux: Uses a separate native eframe (Rust/egui) host, avoiding WebKitGTK compilation.
• Models: Local ASR via vendored Qwen3-ASR, Windows Foundry Local Whisper, sherpa-onnx, plus cloud models you provide.

Who it's for

Anyone who writes a lot of text—especially developers writing commit messages / PR descriptions, or power users who want to dictate detailed prompts for LLMs (ChatGPT, Claude) faster than typing. It's also for privacy-conscious users who want an open-source alternative to subscription dictation tools and who want to use their own API keys and keep their data on their own machine.

Notable caveats

• macOS build is ad-hoc signed (not Apple-notarized), requiring a one-time Terminal command (xattr -cr /Applications/OpenLess.app) to bypass Gatekeeper.
• Linux requires glibc ≥ 2.39 (e.g., Ubuntu 24.04), and there is currently no in-app updater for Linux packages.
• Cloud ASR is the default for quick start; local ASR is available but varies by platform (macOS has bundled Qwen3-ASR, Windows has local options).

Why it matters

It's a fully open-source, local-first alternative to paid SaaS dictation tools. It shifts the power back to the user: you bring your own model credentials, your dictionary and corrections stay on your machine, and it's specifically engineered to help you compose high-quality prompts for other AI systems by speaking rather than typing.