boris-wispr 0.1.0

Voice input engine — hold-to-talk whisper detection, filler removal, @ file tags

Boris Wispr

Voice input engine for Boris TUI. Speak naturally, get clean text.

Features

• Whisper Detection — Works with quiet/whispered speech via Silero VAD sensitivity tuning
• Filler Word Removal — Strips "um", "uh", "ahh", "like", "you know" automatically
• File Tags (@) — Say "at config dot py" and get @config.py — references files in your project
• Hold-to-Talk — Press and hold a hotkey, speak, release — text appears in Boris
• Real-time Transcription — Uses faster-whisper + Silero VAD for low-latency local processing

Install

pip install -e .

Quick Start

# Standalone test
boris-wispr

# With Boris TUI integration
boris-wispr --boris

Architecture

Microphone → VAD (Silero) → Audio Buffer → faster-whisper → Post-processor → Output
                                                              ├─ Filler removal
                                                              ├─ @ tag expansion
                                                              └─ Punctuation cleanup