Talat launches local-only AI notetaking app for Mac

A new Mac application called Talat has launched to offer a local-only alternative for AI-generated meeting notes. Developed by Nick Payne and Mike Franklin, the application captures audio from the computer’s microphone during meetings on platforms like Zoom, Teams, and Meet, transcribing the audio in real time. Unlike cloud-dependent competitors, Talat processes all audio on-device, meaning it does not store audio data on company servers.

The application enters a market currently occupied by Granola, a competitor AI notetaking app valued at $1.5 billion. Payne built Talat to address the privacy trade-offs of sending voice data to third-party cloud providers. To keep the application lightweight and private, the developers packaged Talat as a 20MB download that requires no user account creation or analytics sharing.

The pricing and technical specifications for Talat include:

• Current Price: A one-time purchase of $49 during its pre-release phase.
• Future Price: The price will increase to $99 upon the 1.0 release.
• Trial: A free trial offering 10 hours of recordings for users with M-series Mac computers.
• App Size: A 20MB local installation.

During meetings, Talat transcribes speech and attempts to assign speakers in real time, allowing users to reassign them, take notes, and edit, delete, or split transcript segments. When a meeting finishes, a local LLM generates a summary with key points, decisions, and action items. All notes, transcripts, and summaries remain searchable within the app.

Under the hood, Talat runs local audio AI on Apple’s Neural Engine—Apple’s dedicated hardware for AI processing—using a Swift framework called FluidAudio. The application defaults to the Qwen3-4B-4bit AI model for generating summaries. However, the developers have prioritized user control and configurability. Users can swap the default model for cloud LLM providers, two Parakeet speech-recognition models developed by Nvidia, or Ollama, a tool for running AI models locally.

Payne highlighted this focus on user control, stating: “We’re leaning into configurability and letting users control where their data goes: pick your own LLM, auto-export to [notetaking app] Obsidian, webhooks that push data out when a meeting finishes, an MCP server” (a Model Context Protocol server, which is a standardized way for AI tools to connect to outside data sources).

Why it matters

Talat provides a privacy-focused, local-only alternative to popular AI notetaking apps like Granola, allowing users to keep audio data on their own machines without ongoing subscription fees.