Overview
This mod lets you communicate in Minecraft using your voice instead of typing every message manually. Simply hold the configured voice input key (V by default), speak into your microphone, and release the key to automatically transcribe your speech and send the resulting text through Minecraft's chat.
Features
Push-to-talk voice input: Hold the configured key while speaking, then release it to finish recording and trigger transcription and message sending.
Automatic message sending: The recognized text is sent to Minecraft chat automatically after you release the key, without requiring you to type the message yourself.
Built-in speech recognition model: The mod includes a quantized, 8-bit version of Alibaba's FunASR SenseVoiceSmall speech recognition model, so the recognition model does not need to be downloaded separately.
Designed with size in mind: The quantized model provides a practical balance between model size and recognition quality, helping keep the mod's bundled assets more manageable.
Configurable key binding: V is the default voice input key, and the key can be changed to suit your preferences.
Speech Recognition
The mod uses SenseVoiceSmall, a multilingual speech recognition model from the FunASR ecosystem. SenseVoiceSmall supports Mandarin Chinese, Cantonese, English, Japanese, and Korean, although the recognition results may vary depending on the language, pronunciation, background noise, and technical vocabulary being spoken.
The included 8-bit quantized model is a deliberate compromise between recognition accuracy and distribution size. Recognition is not perfect, and errors may occur, especially with Minecraft-specific terminology, player names, modded item names, or other uncommon words.
Development
This project was developed with assistance(actually 100% AI generated, I only designed the icon) from GitHub Copilot and GPT-5.6 Terra.

