← All projects

ditado — Offline Dictation & Audio Recording

Offline-first Windows application for private AI dictation and meeting recording using local speech recognition. Supports 13 languages + auto-detect with zero cloud dependency.

Role
Creator
Year
2026
Stack
Python · faster-whisper · sounddevice · pyaudiowpatch · ffmpeg · PyInstaller

Context

Cloud-based speech recognition raises privacy concerns, requires internet connectivity, and performs inconsistently for Brazilian Portuguese. Professionals also need an easy way to capture meetings locally without relying on third-party recording services.

Approach

Designed and built a desktop application that runs entirely offline from the Windows system tray.

  • F8 Dictation: record speech, transcribe locally with faster-whisper, and automatically paste the result into any application.
  • F9 Recording: capture microphone and system audio simultaneously using WASAPI loopback with stereo channel separation.
  • Supports 13 languages + automatic language detection.
  • Packaged as a standalone Windows installer using PyInstaller and Inno Setup.
  • MIT licensed and fully open source.
  • Outcome

  • Production-ready Windows release.
  • Near-human transcription quality for Brazilian Portuguese.
  • 100% local processing with no cloud services.
  • Enables private dictation and meeting recording without subscriptions or internet access.