Echosy

Echosy is an macOS desktop app that captures system audio and microphone, transcribes speech in real-time using on-device ASR models (Qwen3-ASR, MLX Whisper, Whisper), and provides system-wide dictation — all running 100% locally. No cloud uploads, no subscriptions, no data collection. Your recordings, transcripts, and models live entirely on your machine.

Live Transcription System audio + microphone captured via ScreenCaptureKit VAD-based chunking with adaptive noise floor — no fixed-interval cuts Real-time segments with timestamps streamed as they are recognized Audio level waveform visualization during recording Auto-save transcripts as Markdown, optional WAV recording Export as .md or .txt, copy to clipboard ASR Model Management 10 models across three backends: Qwen3-ASR (GPU via MPS): 0.6B (~1 GB) and 1.7B (~2 GB) MLX Whisper (GPU via Metal, Apple Silicon only): Small, Large V3 Turbo, Large V3 Whisper (CPU fallback): Tiny, Base, Small, Medium, Large V3 One-click download with real-time progress tracking (bytes downloaded / total) Switch models on the fly — old model is unloaded from GPU memory before loading the new one Delete unused models to reclaim disk space Universal Dictation Cmd+Shift+D to start/stop from any app Speech is transcribed locally and pasted at your cursor Works with CJK text (Chinese, Japanese, Korean) Live waveform and segment history in the Dictation tab Streaming Summary Send transcript to OpenAI, Gemini, or any OpenAI-compatible endpoint Token-by-token streaming with rendered Markdown preview Customizable system prompt (meeting notes, lecture notes, etc.) Auto-saved alongside the transcript Session History Browse all past recordings with duration, size, and date Search, sort, inline rename, and delete Preview transcripts and summaries without leaving the list Load any past session back into the app for review or re-summarization File Transcription Drag & drop WAV, MP3, MP4, MOV, or M4A files Chunked transcription with real-time progress Fully offline — same local ASR models used for live recording Multi-Language UI English, ????, ????, ???, ??? Auto-detects from system locale, switchable in Settings Appearance Dark, Light, and System theme (follows macOS) Adjustable font size (Small / Medium / Large) System Integration System tray with recording/dictation controls and active indicator Global hotkeys: Cmd+Shift+R (recording), Cmd+Shift+D (dictation) Permission checks with one-click links to System Settings Configurable workspace folder for all recordings and transcripts