talat

talat is a privacy-first AI meeting note-taker for macOS and Windows that records, transcribes, labels speakers, and summarises every meeting entirely on your own device — no account, no cloud, no subscription.

What is talat?

talat is the meeting-notes app that refuses the cloud bargain. Built by Yorkshire-based developer Nick Payne and his friend Mike Franklin under Lumikey LTD, it does what Granola, Otter, and Fireflies all outsource: it records both your microphone and system audio, transcribes every word in real time, identifies who said what, and drafts a summary with key points, decisions, and action items — all on your own machine, using your Mac's Neural Engine or Windows GPU. There's no bot to invite, no account to create, and nothing is uploaded unless you deliberately point summaries at a cloud LLM with your own API key. Everything lands in a local database with full-text search, Markdown and PDF export, webhooks, calendar sync, and an MCP server so tools like Claude can query your meeting history. After a TechCrunch profile and a Product Hunt launch in March 2026 that drew 146 upvotes, it ships as a 20MB app with a 10-hour free trial and a one-time licence that covers every future update. The pitch is simple: your meetings are yours, so the software shouldn't rent them back to you.

Key Features

• *100% On-Device AI** — recording, transcription, speaker labelling, and storage all run locally; nothing is uploaded, and the app works offline or fully air-gapped
• *Real-Time Meeting Capture** — dual-stream mic and system audio with echo cancellation, auto-start/stop across Zoom, Teams, Google Meet and more, plus per-channel mute and live level meters
• *AI Meeting Notes** — a summary, action points with owners, and timestamped chapters generated the moment a call ends, on a local model or the cloud LLM of your choice
• *Speaker Identification** — voices are split into distinct speakers as they record and recognised across future meetings via voice prints, with a live presence rail showing who's in the room
• *Live Transcription & Editing** — streaming on-device speech-to-text with custom vocabulary for names and jargon, inline editing, and instant full-text search across every meeting
• *System-Wide Dictation** — press a hotkey and speak into any app; optional LLM cleanup fixes punctuation and filler, with per-app overrides and searchable history
• *Local File Import** — drop in WAV, MP3, M4A, MP4, FLAC, OGG, or AAC recordings and get the full transcription, speaker labelling, and summary pipeline with no upload
• *Screenshot Capture** — grab slides and diagrams mid-meeting with a hotkey, anchored to the transcript timeline and made searchable via on-device text recognition
• *Open Integrations** — Markdown/PDF export, webhooks, automatic folder export, Apple/Google calendar sync, and an MCP server so AI assistants can read and write your meeting history

Who is it for?

• Solo founders, consultants, and freelancers running 5-10 calls a week who are tired of $15-$30/month per-seat subscriptions
• Privacy-conscious professionals in legal, healthcare, and finance whose meeting content must not leave the device
• Journalists, coaches, and creators who turn interviews and client calls into written work and want a private capture layer
• Technical users who want webhooks, MCP, custom summary prompts, and bring-your-own-LLM control
• Small teams on Mac and Windows that want reliable notes without a visible recording bot in client meetings

Who is it NOT for?

• Teams that need shared workspaces, centralised admin controls, SSO, or audit logging
• Anyone on Linux, older Intel Macs, mobile, or browser-only workflows — it simply won't run
• Buyers who need frontier-grade transcription on noisy rooms, overlapping speakers, or heavy accents
• Sales and ops teams that want meetings auto-logged into Salesforce or HubSpot without building webhook bridges
• Casual note-takers who don't care where audio goes — a free cloud tier will do the job