AI-powered note-taking app Granola, valued at $250 million, has become a popular tool among tech industry founders and VCs. But one developer believes there is a need for a private, location-only approach available for a one-time fee and without a subscription. That led to the creation of a new Mac app called Talat.
Yorkshire, England-based developer Nick Payne, a self-described computer scientist, says the idea to build Notetaker’s AI environment came about largely through a series of happy accidents.
“I think Granola is amazing; it’s a great example of what you can do with an Electron app.” [a framework for building desktop applications] it’s been given enough love and care,” he told TechCrunch. “When I tried it for the first time, I was impressed that it was able to record system audio on my Mac without recording video, which was a common task at the time. That led to a lot of research, and a new and poorly documented Apple API was discovered.”
To make it easier to work with that API (Core Audio Taps, which allows developers to tap into Mac audio streaming), Payne decided to create an open source library, AudioTee.
“At the time, I was slowly putting together a set of tools, but I couldn’t find anything that felt like it could stand on its own as a product rather than just a cool tech demo,” Payne said. “The high-end hosting transcription models – the same providers that people like Granola use – are amazing, and it’s visually great to see your speech projected on the screen soon. But it always bothers me that the exchange needs to provide not only my data, but my audio data; my real voice,” he added.
Then he came across a software toolkit called FluidAudio, a Swift framework that enables fully localized, low-cost audio AI for Apple devices. It allows you to run small, fast written models directly on the Mac’s Neural Engine – Apple’s dedicated AI processing hardware.
That was the piece that made Payne realize he could turn his research into a real product — where your audio never leaves your Mac and your documents aren’t stored on another company’s servers.
Talat, formed around Payne’s longtime friend and former collaborator Mike Franklin, is the result of Payne’s interest in the soundscape. The result is a 20MB, one-time purchase that doesn’t require you to create an account or share analytics data with developers. There are no ongoing fees, either.
While other AI note takers may have more bells and whistles, Talat offers an advanced feature set. It captures audio from your computer’s microphone when you’re connected to apps like Zoom, Teams, Meet, and others, and records it in real time. The app tries to assign speakers in real time, but you can reassign them as needed. You can also take notes, and edit, delete, or split segments of the transcript. At the end of the meeting, the local LLM produces a summary with key points, decisions, and action items.
Notes, transcripts, and summaries are all searchable in Talat, too.
In addition to the privacy angle, Payne said the goal is to give users more options.
“We rely on customization and let users control where their data goes: choose your own LLM, automatically export to [notetaking app] Obsidian, webhooks that push data when the session ends, MCP server,” which is a standard way for AI tools to connect to external data sources, “pulling it when needed,” he explained.
Under the hood, the AI is a hybrid — “it’s mostly cobbled together and released behind FluidAudio,” notes Payne, who says he’s doing a lot of hard work. In a nutshell, the app automatically switches to an Al model called Qwen3-4B-4bit, which can run even on modest hardware.
However, users can choose to switch that to any cloud LLM provider of their choice, or they can choose between two variants of Parakeet – speech recognition models developed by Nvidia – or point to Ollama (a tool for running AI models locally), which gives them more control over the experience. Over time, Talat will add support for more built-in options and will have integration with other applications, such as Google Calendar and Notion.
At launch, users with M-series Mac computers (those using Apple processors, starting with the M1) can download the app and try it for free with 10 hours of recording time before making a purchase decision.
Talat is available for $49 while in this pre-release, performance-enhanced version.
When the app reaches release 1.0, the price will increase to $99.
Payne and Franklin opened Talat and planned to keep the core product one-time only going forward.
