LiteSpeak

Overview

LiteSpeak adds system-wide voice dictation to any application. Press a hotkey, speak, and LiteSpeak transcribes your audio and pastes the result directly into whatever window has focus. An optional LLM refinement pass cleans up grammar, punctuation, and filler words before the text lands.

Features

• Global hotkey for system-wide dictation — works in any app
• Approval gates — vocal authorization protocols with configurable timeouts, TTS read-back confirmation, and echo prevention for safe hands-free operation
• 7 LLM providers — GPT-4.1, Claude 4.5, Gemini, local models via LM Studio, and more
• Custom modes with per-mode system prompts, language presets, and model overrides
• Managed local TTS with voice selection and speed control
• 4 STT providers for speech-to-text transcription
• HTTP API on port 7428 for programmatic access and integration
• Optional LLM refinement pass to clean up grammar and punctuation
• Automatic paste into the active window on transcription complete

Installation

Select LiteSpeak during the LiteAISuite install process. After installation, LiteSpeak runs as a background process and registers its global hotkey automatically. It appears in the system tray when active.

Usage

Basic dictation:

1. Click into any text field in any application.
2. Press the configured hotkey (default: Alt+Shift+S).
3. Speak. A small indicator appears to confirm recording is active.
4. Press the hotkey again (or pause) to stop recording.
5. The transcribed text is pasted at the cursor position.

LLM refinement: Enable Refine with LLM in settings. After transcription, LiteSpeak sends the raw text to your configured LLM endpoint (local or API-based) and pastes the cleaned result instead of the raw transcript. Adds a small delay depending on model speed.

Hotkey customization: The global hotkey is configurable in Settings. Choose any combination that doesn't conflict with your other tools.

Configuration

Open the LiteSpeak settings window from the system tray icon. Key options:

• Hotkey — The key combination to start/stop recording
• Transcription model — Choose from available local Whisper model sizes (tiny, base, small, medium)
• LLM refinement — Toggle on/off; set endpoint URL and model if using a local LLM
• Paste method — Clipboard paste (default) or simulated keystrokes for apps that block clipboard input
• Audio input device — Select which microphone to use

Requirements

• A working microphone connected to the system
• No additional requirements beyond the base Lite Suite install (transcription runs locally via bundled Whisper)

Troubleshooting

The hotkey doesn't trigger recording. Another application may have claimed the same key combination. Open LiteSpeak settings and assign a different hotkey, then restart LiteSpeak.

Transcription is slow or inaccurate. Try a smaller Whisper model (tiny or base) in Settings for faster results, or a larger model (medium) for better accuracy. Model size is a quality/speed tradeoff.

Text doesn't paste into the target app. Some applications (notably certain browser address bars and admin-elevated windows) block clipboard paste. Switch the Paste method setting to Simulated keystrokes and retry.