Za-Killer Vosk Fabric

Za-Killer Vosk Fabric

A voice-reactive Minecraft challenge mod.

Sound expands your vision.
Silence makes the world dark.
Saying “자” can kill you.

This mod was inspired by Hwiyong’s voice challenge video and is designed for singleplayer challenge runs.

Important: Vosk model required

This Modrinth file only contains the mod jar.

To use voice recognition, you must separately download the Vosk Korean model:

vosk-model-small-ko-0.22

Official Vosk model page:

https://alphacephei.com/vosk/models

Direct model download:

https://alphacephei.com/vosk/models/vosk-model-small-ko-0.22.zip

Download the model ZIP, extract it, and make sure the extracted folder name is exactly vosk-model-small-ko-0.22.

After downloading, extract the ZIP and place the model folder like this:

mods/
  za-killer-vosk-fabric-26.1-26.1.2-1.0.0.jar
  za-killer-vosk/
    vosk-model-small-ko-0.22/

If the model folder is missing, the mod may load, but voice recognition will not work.

Requirements

• Minecraft 26.1 / 26.1.1 / 26.1.2
• Fabric
• Java 25
• Fabric API
• Microphone

Optional:

• Mod Menu

Gameplay

The mod uses your microphone volume to control vision distance.

• No sound: very limited vision
• Louder sound: wider vision
• Trigger word detected: death

Default trigger text: 자
Default difficulty: Hard

Easy difficulty

The player dies only when the configured trigger text is recognized almost by itself.

Example:

If the trigger text is 자, the player mainly dies when 자 is recognized as a standalone result.

Hard difficulty

The player dies when the configured trigger text appears anywhere in the recognition result.

Example:

If the trigger text is 자, results such as these can kill the player:

• 자
• 자작나무
• 감자
• 시작하자

Hard is the default difficulty.

Configuration

Press F8 in-game to open the Za-Killer settings screen.

You can configure:

• Minimum vision distance
• Maximum vision distance
• Silence threshold
• Loud voice threshold
• Vision change speed
• Trigger text
• Difficulty
• HUD mode

Notice

Korean speech recognition is powered by Vosk and may not be perfect.

Misrecognition can happen.
In this mod, occasional misrecognition is part of the challenge.

This mod was fan-made with the assistance of ChatGPT.
ChatGPT was used for planning, feature design, code generation, debugging, documentation, and release text drafting.

Source and licenses

Source code and third-party license notices are available on GitHub.

This project uses:

• Vosk API
• Vosk Korean speech recognition model
• Java Native Access, JNA
• Fabric API