Thoth

Thoth is a macOS application that records and transcribes both microphone and system audio simultaneously, allowing users to capture the full conversation of meetings, calls, or lectures without relying on third‑party routing tools. The software runs entirely on the local machine, using WhisperKit and CoreML for transcription and speaker diarization, and can export results in PDF, Markdown, RTF, or JSON formats with timestamps and color‑coded speaker labels.

Targeted at individuals who handle confidential information, Thoth keeps all data on the device, never uploading audio to the cloud. It offers on‑device AI summarization with a selection of local models ranging from 1.9 GB to 6.8 GB, and optionally supports external API keys for services such as OpenAI, Anthropic, or Google, which are stored securely in the Apple Keychain. The app is built with native SwiftUI, supporting Dark Mode and running on Apple Silicon for accelerated performance.

Performance benchmarks on an M2 MacBook Pro show transcription of a 42‑minute recording in about 3.3 minutes (12.7× real‑time) and speaker diarization in under eight seconds for two speakers, with support for up to eight speakers and automatic language detection for 99 languages. The tool is provided as a free trial with a paid Pro option, emphasizing privacy‑by‑default and offline operation.