Open-Less/openless

Hold a key, speak, release — AI-polished text appears at your cursor in any app. Open-source voice input for macOS & Windows. (按住快捷键说话，松开即得润色后的文字)

What it solves

OpenLess is a cross-platform voice-input application that eliminates the friction between spoken thoughts and clean, usable text. Unlike standard dictation tools that provide verbatim transcripts, OpenLess uses AI to polish, structure, and format spoken input into professional text or detailed AI prompts, inserting them directly at the cursor in any application.

How it works

Users hold a global hotkey to record audio, which is then processed through a pipeline of Automatic Speech Recognition (ASR) and Large Language Model (LLM) polishing. The system supports both cloud-based ASR (Volcengine, OpenAI Whisper) and local ASR (Qwen3-ASR, Windows Foundry Whisper). The polished text is then streamed character-by-character to the active text field via a native OS insertion protocol, with a clipboard fallback if the application blocks direct insertion.

Who it’s for

  • AI Power Users: People who want to dictate complex, structured prompts for LLMs like ChatGPT, Claude, or Cursor without typing.
  • Writers and Professionals: Those drafting emails, specifications, or Slack messages who prefer speaking but want the output to be grammatically correct and organized.
  • Developers: Users needing to capture code comments, commit messages, or PR descriptions quickly.

Highlights

  • AI-Prompt Mode: Specifically designed to reshape loose speech into structured prompts with context and constraints.
  • Style Pack Marketplace: A community-driven system allowing users to create, install, and switch between custom system prompts (style packs) via hotkeys.
  • Streaming Insertion: Reduces perceived latency by flowing text to the cursor in real-time as it is polished.
  • BYO-Credentials: Supports bringing your own cloud credentials for ASR and LLM providers (e.g., DeepSeek, OpenAI, Anthropic) to avoid vendor lock-in.
  • Local-First: Open-source architecture where data and dictionaries stay on the user's machine.