InputCli

Terminal-based voice input tool with AI transcription and smart post-processing.

Say something → text appears at your cursor. Transcription, correction, and formatting are fully transparent.

Features

• Press-and-hold voice input — configurable global hotkey (F13/F5/CapsLock)
• ElevenLabs Scribe transcription with Chinese language support
• Claude Haiku post-processing — filler removal, hotword correction, punctuation
• Auto-injection — text is pasted into the active window via clipboard
• TUI interface — real-time status, volume meter, history, hotword management
• Sound feedback — synthesized event sounds with theme support
• SQLite history — searchable transcription log with in-memory LLM context

Quick Start

# Install with uv
uv sync
uv run inputcli

# Or install with pip
pip install -e .
inputcli

Configuration

Config file: ~/.inputcli/config.yaml (auto-created on first run).

Set your API keys via environment variables:

export ELEVENLABS_API_KEY="your-key"
export ANTHROPIC_API_KEY="your-key"

CLI Options

inputcli --help
inputcli --hotkey f5         # override hotkey
inputcli --language en       # override language
inputcli --no-postprocess    # skip LLM post-processing
inputcli --debug             # enable debug logging

Architecture

HotkeyListener → AudioRecorder → Transcriber → PostProcessor → Injector
     ↕                                              ↕
  Pipeline (FSM)  ←→  SoundPlayer / HistoryStore / Config
     ↕
  TUI (Textual)

Requirements

• Python 3.11+
• macOS or Linux
• PortAudio (for sounddevice)
• Microphone access