A new Mac app called Talat wants to prove that AI meeting notes don’t need to leave your computer. Built by Yorkshire-based developer Nick Payne and his longtime friend Mike Franklin, Talat is a local-first alternative to cloud-based notetakers like Granola, TechCrunch AI reports.
The pitch is simple: your audio, transcripts, and summaries stay on your machine. No account required. No analytics. No subscription. Just a one-time purchase.
How It Started
Payne’s journey to Talat began with curiosity about how Granola managed to record system audio on a Mac without capturing video. “I think Granola is awesome; it’s a shining example of what you can do with an Electron app given enough love and care,” Payne told TechCrunch AI.
That curiosity led him down a rabbit hole into Apple’s Core Audio Taps API, a relatively new and poorly documented interface for tapping into a Mac’s audio streams. He built an open source library called AudioTee to make it easier to work with. But the real breakthrough came when he discovered FluidAudio, a Swift framework that runs small transcription models directly on Apple’s Neural Engine.
What nagged Payne about existing tools was the privacy trade-off. “It always nagged me that the trade-off required providing not just my data, but my audio data; my actual voice,” he said.
What Talat Does
The app weighs just 20MB and offers a focused feature set:
- Real-time transcription from your microphone during Zoom, Teams, Meet, and other meeting apps
- Speaker detection with the ability to reassign speakers manually
- Live note-taking alongside the transcript, with editing and splitting capabilities
- Local AI summaries generated after meetings with key points, decisions, and action items
- Full search across notes, transcripts, and summaries
For summarization, Talat defaults to Qwen3-4B-4bit, a model small enough to run on modest hardware. But users can swap in any cloud LLM provider, choose between two Nvidia Parakeet speech-recognition models, or point it at Ollama for local inference.
What stands out here is the configurability angle. Payne described it as letting users “pick your own LLM, auto-export to Obsidian, webhooks that push data out when a meeting finishes, an MCP server to pull it on demand.” That level of control is rare in consumer-facing meeting tools.
Pricing and Availability
Talat requires an M-series Mac (M1 or later). New users get 10 hours of free recordings to test things out. During the pre-release period, it’s $49 as a one-time purchase. When version 1.0 ships, the price goes up to $99.
Payne and Franklin are bootstrapping the company and plan to keep the one-time purchase model. No venture funding, no subscription treadmill.
Future updates will add more built-in model choices and integrations with Google Calendar and Notion, according to TechCrunch AI.
Why This Matters
The AI meeting notes space has exploded, but almost every player requires sending your audio to the cloud. Granola, valued at $250 million, is the current darling among tech founders and VCs. Talat isn’t trying to compete feature-for-feature. It’s targeting a specific crowd: people who care where their voice data ends up.
The one-time purchase model is also a deliberate counterpoint to the subscription fatigue hitting the productivity software market. Whether a bootstrapped two-person team can sustain development on one-time sales is the obvious question, but at $49-$99 per user with zero infrastructure costs for cloud processing, the math might actually work.
For more details on Talat’s launch and technical underpinnings, check the full report at TechCrunch AI.