# OpenLess

> Open-source voice input app that turns held-hotkey speech into AI-polished text inserted at your cursor on macOS, Windows, Linux and Android.

OpenLess is an open-source voice-input application from the Open-Less organization. You hold a global hotkey, speak, and the app transcribes and polishes the speech, then inserts the text at the cursor in any app. The README lists macOS, Windows, Android and a native Linux host, and the latest listed release is v2.0.1-tauri.

## What It Is

OpenLess is a dictation and voice-to-text tool that sits between speech and the active text field. The pipeline is speech, transcript, cleanup, then insertion at the cursor, with a clipboard fallback if insertion is blocked. Its model only reshapes text: it does not answer questions or run tasks, so a spoken question comes back as a cleaned-up question.

A distinguishing mode is the AI-prompt mode, which turns loose speech into a structured, context-rich prompt that can be pasted into tools such as ChatGPT, Claude or Cursor. Four output modes are documented: raw, light polish, structured and formal, plus a translation hotkey.

## Style packs and dictionary

Style packs are named output styles, each with its own custom system prompt and switchable by hotkey. Users can install community packs from an in-app Marketplace or publish their own; uploads are moderated.

A dictionary holds proper nouns and names. Entries are sent as hotwords to ASR providers that support them and as hints to the polish model. An experimental, opt-in feature can suggest dictionary additions from your corrections, and each needs confirmation. An opt-in cursor-context feature lets the polish model read text around the cursor on macOS, Windows and Android.

## Providers and local data

Users bring their own credentials. Documented cloud ASR options include Volcengine, iFlytek, Alibaba Cloud Bailian, StepFun, ElevenLabs Scribe and OpenAI-compatible endpoints. Local ASR options include bundled Qwen3-ASR on macOS and experimental Windows variants. Polish providers include Ark, DeepSeek, OpenAI, Google Gemini and OpenRouter, plus any OpenAI-compatible endpoint. Credentials are stored in the OS vault (Keychain, Credential Manager or Linux keyring).

## Setup path

Installers are distributed through GitHub Releases, and a Homebrew Cask is available for macOS. Because the macOS build is ad-hoc signed rather than notarized, the README says to run an xattr command once. First launch asks for microphone and accessibility permissions plus ASR and polish credentials. Source builds use Rust and a Tauri 2 host with a React/TypeScript frontend, and Linux uses a native egui host. The project is released under AGPL-3.0-only.

## Features
- Global hotkey push-to-talk and toggle recording
- AI-prompt (structured) output mode
- Raw, light polish, structured and formal output modes
- Translation hotkey
- Streaming character-by-character insertion with paste fallback
- Clipboard fallback when insertion is blocked
- Style packs with custom system prompts and community Marketplace
- Dictionary hotwords with opt-in learning from corrections
- Opt-in cursor context for polishing
- Selection-ask voice Q&A panel
- Cloud and local ASR options
- Bring-your-own ASR and polish providers
- Credentials stored in OS vault
- Multilingual UI
- In-app auto-update and opt-in Beta channel

## Integrations
Volcengine, iFlytek, Alibaba Cloud Bailian, StepFun, Tencent Cloud, Zhipu GLM-ASR, ElevenLabs Scribe, OpenAI, Google Gemini, DeepSeek, OpenRouter, SiliconFlow, Groq, Apple Speech, Homebrew

## Platforms
WINDOWS, MACOS, LINUX, ANDROID, IOS, WEB, API, VSC_EXTENSION

## Pricing
Open Source

## Version
v2.0.1-tauri

## Links
- Website: https://github.com/Open-Less/openless
- Documentation: https://github.com/Open-Less/openless/blob/beta/USAGE.md
- Repository: https://github.com/Open-Less/openless
- EveryDev.ai: https://www.everydev.ai/tools/openless