Two local-first, no-bot tools. heed is a free MIT-licensed CLI with channel-based speaker diarization for Linux and macOS. Clearminutes is a polished cross-platform desktop app with a live privacy network monitor.
Two local-first tools: different shape, same privacy stance.
heed is a genuinely local-first open-source project; MIT licensed, $0 forever, with channel-based speaker diarization that flags overlapping speech via pyannote. It runs on Linux and macOS as a CLI-installed floating panel, captures Zoom, Meet, Teams, and Discord audio without a bot, and generates summaries through Ollama. Clearminutes matches its local stance and adds Windows support today (heed's Windows is on the roadmap), a live transcript view in a full desktop window, and a built-in privacy network monitor. Choose Clearminutes for the polished product; choose heed if you want a fully auditable MIT codebase and channel-based overlap detection.
| Clearminuteslocal-first | heedcloud-first | |
|---|---|---|
| Capture & privacy | ||
| Meeting bots | No | No |
| Audio stays on device | Yes | Yes |
| On-device transcription | Yes (on-device) | Yes (faster-whisper) |
| Local processing | Yes (on-device models) | Yes (faster-whisper + pyannote + Ollama) |
| Privacy network monitor | Yes (live, in-app) | No |
| No account required | Yes (local mode) | Yes |
| Transcription | ||
| Live transcript while recording | Yes | Yes (real-time as you record) |
| AI & summaries | ||
| AI summaries | Yes (local or BYO keys) | Yes (local Ollama) |
| Local LLM for summaries | Yes (Gemma, built-in) | Yes (Ollama, 14 model options) |
| Bring-your-own cloud AI keys | Yes (Claude, Groq, OpenRouter) | No (no API keys needed) |
| Custom templates | Yes | Yes (general, sales, 1-on-1, retro, custom) |
| Transcription | ||
| Speaker identification | Yes (Pro) | Yes (voice memory: rename once, remembered) |
| Speaker diarization | Yes | Yes (pyannote 3.1, channel-based) |
| Workflow | ||
| Action items + decisions | Yes + TickTick sync (Pro) | Not specified |
| Platform & performance | ||
| Cross-platform | macOS + Windows + Linux | Linux + macOS (Windows coming soon) |
| GPU acceleration | GPU-accelerated | Yes (CUDA, MPS, CPU auto-split) |
| Audio & export | ||
| Audio file import | Yes (Pro) | Not specified |
| Works with | Any audio source | Zoom, Meet, Teams, Discord, YouTube, podcasts |
| Capture & privacy | ||
| Works offline | Yes (local models) | Yes (no internet after setup) |
| Integrations | ||
| Integrations | TickTick, Outline, PDF | None (Calendar, Markdown/Obsidian/Notion on roadmap) |
| Claude Desktop MCP connector | No | No |
| Workflow | ||
| Auto call detection | Yes (Pro) | Yes (detects Zoom/Meet/Teams/Discord) |
| Cross-meeting search | Yes | Not specified |
| Pricing | ||
| Free tier | Yes | Yes (unlimited, MIT) |
| Pricing model | Free + Pro £12/mo + Lifetime £249 | Free ($0, MIT) |
| Funding | Bootstrapped | Solo project (isjunrod) |
| Transcription | ||
| Language support | 6+ languages | English + Spanish UI; multi-language transcription |
| Workflow | ||
| Guided onboarding | Yes | Yes (setup wizard + in-app tour) |
| Commercial support | Yes | No (community only) |
| Active maintenance | Yes | Yes (actively maintained, solo developer) |
| Open source | No | Yes (MIT) |
Local transcription for unlimited meetings, no bot, no cloud. Pro only if you want speaker ID and cross-device sync.
Last updated: 2026-06-22. Compared from each tool's public pricing and feature data (see sources), as of the last-updated date.
Sources: heed GitHub repository (2026-06-22), heed README (2026-06-22).
Clearminutes is our own product; we've kept the comparison fair.