Speech-2-TTS Mod by NocturnalLyfe

Transform your voice chat into hilarious text-to-speech! Uses completely offline speech recognition.

Features

• Real-time voice -> text -> TTS conversion
• Your actual voice is muted when STT is enabled (only TTS is broadcast)
• Toggle on/off with a single keypress
• Text appears in game chat for all players to see
• Flexible model support - use small, medium, or large models
• No performance impact

Installation

Automatic (Mod Manager)

1. Install via Thunderstore Mod Manager or r2modman
2. Launch the game

Manual

1. Install BepInEx if not already installed
2. Extract the mod folder to REPO/BepInEx/plugins/
3. Launch the game

Usage

Controls

• F7 - Toggle Speech-to-Text mode ON/OFF

Modes

STT Enabled (Default):

• Your voice is converted to text
• Text is broadcast via in-game chat as TTS
• Your actual voice is muted (others only hear TTS)
• Perfect for robot voice fun!

STT Disabled:

• Normal voice chat
• Your real voice is transmitted
• No text-to-speech conversion

How It Works

1. Speak into your microphone
2. Vosk speech recognition transcribes your voice
3. Text is sent to the game's chat system
4. Game's TTS speaks your message to all players
5. Everyone sees and hears your transcribed speech!

Configuration

No configuration needed! Just press F7 to toggle.

Using Different Models

The mod automatically detects which model you have installed (in priority order):

1. vosk-model-small-en-us-0.15 (~40MB) - Default, fastest, good accuracy
2. vosk-model-en-us-0.22 (~1.8GB) - Better accuracy, slower
3. vosk-model-en-us-0.42-gigaspeech (~2.3GB) - Best accuracy, slowest

To Use a Different Model:

1. Download your preferred model from https://alphacephei.com/vosk/models
2. Extract to BepInEx/plugins/NocturnalLyfe-Speech2TTS/model/
3. Keep the original folder name (don't rename it)
4. The mod will automatically detect and use it!

Tip: You can have multiple models installed - the mod will use the first one it finds in the priority order above.

Warning: Changing to larger model will cause longer launch of REPO and requires a stronger device to run.

Troubleshooting

Mod not working:

• Check BepInEx/LogOutput.log for errors
• Ensure all DLLs are in the plugin folder (libvosk.dll, libgcc_s_seh-1.dll, libstdc++-6.dll, libwinpthread-1.dll)
• Verify at least one model folder exists in the model/ directory

Low accuracy:

• Speak clearly and at a moderate pace
• Check microphone quality
• Consider upgrading to vosk-model-en-us-0.22 or vosk-model-en-us-0.42-gigaspeech for better accuracy

No audio transmission:

• This is normal when STT is enabled! Your voice is muted and replaced with TTS
• Press F7 to disable STT for normal voice chat

Wrong model being used:

• Check the log file to see which model was loaded
• Remove unwanted models from the model/ folder
• Models are loaded in priority order (small -> normal -> gigaspeech)

Technical Details

• Default Model: vosk-model-small-en-us-0.15 (~40MB)
• Speech Recognition: Offline Vosk engine
• Sampling Rate: 48kHz
• Language: English (US)
• Supported Models: Any Vosk English model compatible with the API

Credits

• Mod by NocturnalLyfe
• Vosk Speech Recognition: https://alphacephei.com/vosk/
• BepInEx: https://github.com/BepInEx/BepInEx

Support

Report issues on GitHub or the Thunderstore page. https://github.com/NocturnalLyfe/REPO-Speech2TTS

Changelog

See CHANGELOG.md for version history.

Changelog

[1.4.0] - 11-19-2025

Added

• Added Pre-roll capture to add extra audio to speech to better capture first words spoken
• Loading Popup for when loading a model

Changes

• Pre-roll capture Slider for Model
• Asynchronous loading of Models, now letting users continue to play while models load
• Fixed bugs when loading and exiting Settings Menu