talat

Talat is a localized transcription tool for meeting recording, speech dictation, and audio transcription.

All speech content is recognized and stored on the user’s computer without being uploaded to any external server. The software offers three core use cases: meeting recording, real-time dictation, and audio import, all sharing the same proprietary transcription engine. talat In meeting mode, it supports dual-stream audio capture and simultaneous real-time note generation, can connect to Apple Calendar and Google Calendar, automatically prepares recordings based on schedules, and can recognize and automatically initiate calls from Zoom, Teams, Meet, etc. Real-time transcription runs entirely on the device, with speaker recognition continuously learning different voice features. Built-in or user-selectable cloud-based large models can output meeting summaries, to-do lists, and chapter divisions. talat For dictation, users can press a global hotkey to input speech in any application; the speech model is consistent with the meeting, and a large model can be selected to correct punctuation and interjections. Text is automatically pasted into the current window, and the entire dictation history is searchable locally. The import function supports WAV, MP3, M4A, MP4, FLAC, OGG, AAC, and other formats, using the same transcription, speaker, and custom dictionary workflow, and provides a progress bar and complete note processing. The application is compatible with various calendar and meeting software and AI models, and can export to Markdown, Obsidian, PDF or connect via Webhook, making it suitable for professional users who value data sovereignty and offline availability.