Download Vosk models separately and place them here. Models are not included in this repository.
A CLI-first scanner audio transcription tool built with Vosk.
ScannerScribe is a local scanner-audio transcription project. The current version is a command-line transcriber that listens to a selected Windows audio input, transcribes speech with Vosk, applies scanner-specific cleanup rules, and saves transcripts.
The long-term goal is a GUI tool that can show raw and cleaned scanner text, detect keywords, and send local or email notifications.
The long-term goal is to build this into a user-friendly scanner monitoring tool with:
- a GUI
- raw and cleaned transcript windows
- selectable audio input
- configurable scanner correction rules
- keyword alerts
- desktop notifications
- optional email notifications
- optional WAV/audio logging for review
This project is being built in stages. The current focus is a stable CLI transcriber.
This tool is intended for lawful monitoring, accessibility, logging, research, and personal awareness.
Do not use this tool to assist in committing crimes, avoiding law enforcement, interfering with emergency services, or violating the terms of any audio provider.
If using Broadcastify or another streaming service, review their terms of service before automating access, recording, or redistribution.
- Local speech-to-text transcription using Vosk
- Works without sending audio to a cloud API
- Selectable Vosk model path
- Selectable audio input device
- Configurable sample rate
- Timestamped transcript output
- Optional confidence score per transcription line
- Scanner-specific correction rules
- Works with Windows audio inputs such as:
- microphone
- Stereo Mix
- virtual audio cable
- loopback-style routing, depending on setup
- Interactive startup prompts
- Filtered list of usable input devices
- Session name prompt
- Cleaner filename generation
- Optional WAV recording
- Separate raw and cleaned transcript files
- Better correction-rule management
- Keyword detection
- Local desktop notifications
- Audio device dropdown
- Model selection
- Sample-rate selection
- Live cleaned transcript window
- Live raw transcript window
- Correction-rule editor
- Keyword list editor
- Start/stop buttons
- Save-location selection
- Keyword detection
- Desktop notification
- Sound alert
- Email notification
- Possibly SMS or push notification later
The basic pipeline is:
Audio source
↓
Windows audio input
↓
Python sounddevice stream
↓
Vosk speech recognizer
↓
Raw transcript
↓
Scanner cleanup rules
↓
Clean transcript file