Repository navigation
Releases: heidurrus/dictum
Releases · heidurrus/dictum
Release list
v5.2
Full Changelog: v5.1...v5.2
v5.1
v4.0.5
v4.0.0
v3.0.0
v2.3.0
Requirements Workbench for Mac
What's Changed
- Feat/increment 0 hardening by @heidurrus in #1
New Contributors
- @heidurrus made their first contribution in #1
Full Changelog: v1.0...v2.0
What's Changed
- Feat/increment 0 hardening by @heidurrus in #1
New Contributors
- @heidurrus made their first contribution in #1
Full Changelog: v1.0...v2.0
Contributors
heidurrus
Assets 4
v1.0
First public release for Windows.
A local desktop app for transcribing audio using GigaAM — runs entirely on your machine, no cloud, no subscriptions.
Supports Russian, English, and 70+ other languages.
What's included
- All GigaAM model variants (v1/v2/v3 CTC and RNNT, multilingual up to 600M params)
- Automatic longform transcription — no 25-second limit
- Speaker diarization — color-coded "who said what" with timestamps
- Word-level timestamps
- Built-in call recorder — captures system audio + microphone, works with Teams, Discord, etc.
- CPU and GPU inference (NVIDIA CUDA and Apple Silicon MPS)
- Settings UI — paste your HuggingFace token directly in the app, no config files
Installation
Download GigaAM-Transcriber-Setup.exe below and run it. Python 3.10 or newer must be installed first — grab it from
https://www.python.org/downloads/ and check "Add Python to PATH" during setup. The installer handles everything else.
Speaker diarization and longform transcription require a free HuggingFace token — the app will walk you through it in
Settings.
Requirements
- Windows 10/11
- Python 3.10+
- Internet connection on first run (model weights download automatically, ~500MB, cached after that)