# Echosy

> Private, offline audio transcription and dictation app for macOS.

Echosy is a macOS desktop application that captures system audio and microphone input, transcribes speech using on-device AI models, and provides AI-powered summaries and universal dictation — all without any data leaving your Mac.

## Key Facts

- **Platform:** macOS 14+ (Sonoma and later)
- **Architecture:** Apple Silicon and Intel supported
- **Privacy:** 100% on-device processing. No audio data ever leaves your Mac.
- **Offline:** Works without internet. No cloud dependency for core features.
- **ASR Models:** 10 on-device models across 3 backends (Qwen3-ASR, MLX Whisper, Whisper)
- **Languages:** 24+ transcription languages, 5 UI languages (English, Chinese Simplified, Chinese Traditional, Japanese, Korean)
- **Price:** Free tier available. Pro is a one-time $24.90 USD purchase. Enterprise for commercial use.

## Core Features

- **Live Transcription:** Real-time speech-to-text with timestamps from system audio and microphone
- **Universal Dictation:** System-wide dictation with customizable hotkey (Cmd+Shift+D). Text pastes at cursor in any app.
- **System Audio Recording:** Captures audio from any macOS application via ScreenCaptureKit
- **AI Summaries:** Streaming summaries via OpenAI, Gemini, Ollama, or any compatible API
- **Text Enhancement:** Auto-translate, fix punctuation, or apply custom prompts to transcripts
- **Chat with AI:** Ask questions about your transcript and get instant answers
- **File Transcription:** Drag & drop WAV, MP3, MP4, MOV, M4A and 20+ other formats
- **Session History:** Browse, search, and replay past recordings
- **Export:** Markdown, TXT, SRT, VTT, DOCX, PDF

## Privacy Details

- All audio capture, transcription, and dictation happen entirely on-device
- AI models (Whisper, Qwen3-ASR) run locally on your Mac's GPU
- No analytics, telemetry, or tracking
- Optional AI summaries can use local LLMs (Ollama) for complete offline operation
- License validation transmits only the license key — no audio or usage data
- Recordings stored locally in user-chosen workspace folder

## On-Device ASR Models

| Model | Size | Languages | Speed |
|-------|------|-----------|-------|
| Qwen3-ASR 0.6B | ~1 GB | 52 | Fast |
| Qwen3-ASR 1.7B | ~2 GB | 52 | Moderate |
| MLX Whisper Small | ~500 MB | 99+ | Very Fast |
| MLX Whisper Large Turbo | ~2 GB | 99+ | Fast |
| MLX Whisper Large v3 | ~3 GB | 99+ | Moderate |
| Whisper Tiny | ~75 MB | 99+ | Very Fast |
| Whisper Base | ~140 MB | 99+ | Fast |
| Whisper Small | ~460 MB | 99+ | Moderate |
| Whisper Medium | ~1.5 GB | 99+ | Slow |
| Whisper Large v3 | ~3 GB | 99+ | Very Slow |

## Pricing

- **Free:** $0 — 15 min recordings, Qwen3-ASR 0.6B, 3 summaries/day, basic exports (MD, TXT)
- **Pro:** $24.90 one-time — 4 hour recordings, all 10 models, unlimited summaries, file transcription, advanced exports (SRT, VTT, DOCX, PDF), up to 3 devices
- **Enterprise:** Contact sales — required for business, team, or commercial use

## Links

- Website: https://echosy.org
- Privacy Policy: https://echosy.org/privacy
- Terms: https://echosy.org/terms
- Discord: https://discord.gg/hnakzMYh
- Contact: inquiry@echosy.org