Talat: New Mac AI Notetaker Prioritizes Local Privacy

As AI-powered meeting assistants like the $1.5 billion-valued Granola gain traction, a growing segment of users is expressing concerns over the privacy of their voice data. Addressing this demand for local-only processing, developer Nick Payne has launched Talat, a new Mac application that keeps all recordings and transcripts on the user’s machine.

Built alongside longtime collaborator Mike Franklin, Talat differentiates itself by eschewing the subscription model common among productivity tools. Instead, the developers have opted for a one-time purchase fee, positioning the software as a private, permanent utility rather than a service-based platform.

Engineering a Local Solution

The project began as an exploration of Apple’s audio capabilities. Payne, based in Yorkshire, England, initially developed an open-source library called AudioTee to better interact with Apple’s Core Audio Taps API. This allowed him to record system audio without capturing video, a technical hurdle that served as the foundation for the app.

The breakthrough for a standalone product arrived with FluidAudio, a Swift framework that enables low-latency AI transcription directly on the Mac’s Neural Engine. By utilizing this hardware, Talat processes audio locally, ensuring that voice data never traverses a company server.

Features and Customization

While minimalist compared to some cloud-heavy competitors, Talat provides a robust suite of meeting tools:

• Real-time transcription for platforms like Zoom, Teams, and Google Meet.
• Manual and automated speaker assignment.
• Editable transcripts with options to split or delete segments.
• Post-meeting summaries generated by local Large Language Models (LLMs).
• Full-text searchability for all stored notes.

The app defaults to the Qwen3-4B-4bit model for summarization, but Payne emphasizes that the tool is built for user control. “We’re leaning into configurability,” Payne noted. Users can point the application toward their own LLMs, utilize Nvidia’s Parakeet speech-recognition models, or integrate with local tools like Ollama.

Pricing and Availability

Talat is currently available for users with Apple Silicon (M-series) Macs. The developers are following a bootstrap approach, maintaining a clear focus on the one-time purchase model to avoid ongoing fees.

• Trial: Users can record up to 10 hours of meetings for free.
• Pre-release Price: The app is currently listed for $49.
• Full Launch: The cost is scheduled to rise to $99 upon the release of version 1.0.

Looking ahead, the team plans to expand integrations to include services like Notion and Google Calendar, while maintaining the core philosophy that user data should remain private and under the user’s direct control.