Summary
Generate an AI-powered summary of your transcript. Summaries stream in real time and are saved as Markdown inside the recording's .echo archive. You can run summaries against an on-device model (private, no network) or an OpenAI-compatible cloud provider.
| Feature | Tier |
|---|---|
| Generate summary — on-device (Gemma 4 / Qwen 3.5)(private, no API key) | free |
| Generate summary — cloud LLM (OpenAI / Gemini / Claude / Groq / …)(3/day free, unlimited Pro) | free |
| Auto-summary on recording stop | free |
| Generate directly from Recording History (no need to load) | free |
| Auto-generated subject line (short title) | free |
| Markdown preview / raw toggle | free |
| Stop mid-generation | free |
| Custom summary prompt | pro |
| Export summary (md, txt) | free |
| Export summary (srt, vtt, docx, pdf) | pro |
Generate from Recording History
In the Sessions tab, any recording without a summary shows a Generate Summary button. Click it to start streaming a summary in place — the preview updates live, and the recording's subject refreshes automatically when the summary finishes. No need to load the recording into the editor first.
On-device vs. Cloud
Echosy ships with built-in MLX support on Apple Silicon, so you can summarize without sending anything off your Mac.
- On-device (Gemma 4 E4B, Qwen 3.5 4B) — downloads once, ~4–5 GB, no API key, no rate limit, works offline. Recommended for anyone handling private content.
- Cloud (OpenAI-compatible) — faster on longer transcripts, uses your configured provider. Subject to free-tier daily limits.
Custom Summary Prompt
Pro users can customize the prompt in Settings → AI / LLM. This controls the summary structure. Examples:
- "Create meeting minutes with action items and decisions"
- "Write a brief executive summary in bullet points"
- "Extract key takeaways and follow-up tasks"