Skip to content

🚀 Getting Started

🛠 Installing Requirements

UnityNeuroSpeech requires several things to be installed before using it in Unity. Here what you need to install:

1. Ollama

Ollama is a platform for running large language models (LLMs) locally. You can use models like DeepSeek, Gemma, Qwen, etc. Note that small models might affect accuracy and context understanding, but big models response can take a long time.

Install it from the official website.

Then you need to download LLM model with this command: 

ollama pull modelname

For quick test I recommend to download qwen2.5:3b - it responds very fast.

2. STT

Whisper — Speech-To-Text model that transcribes and translates audio with high accuracy.

You need to download Whisper model from here.

For quick test I recommend to download ggml-base.bin.

3. TTS

UV — a modern, ultra-fast Python package and environment manager. It replaces traditional tools like pip. Coqui TTS(runs XTTS) uses UV to simplify installation and allows running the TTS command directly, without manual Python setup.

Coqui XTTS — a Text-To-Speech model that can generate speech in any custom voice you want: Chester Bennington, Vito Corleone (The Godfather), Cyn (Murder Drones) or any other.

 

Install UV with this command:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Then install Coqui TTS:

uv tool install --python 3.11 "coqui-tts==0.27.2"

You can try to install latest Coqui TTS version, but I can't guarantee that it will work.

4. Unity deps

UnityNeuroSpeech uses UniTask for... I think everyone knows what UniTask is. You can install it from the official repository.

Or via UPM:

https://github.com/Cysharp/UniTask.git?path=src/UniTask/Assets/Plugins/UniTask

5. UnityNeuroSpeech

Now you can finally import UnityNeuroSpeech to your Unity project.

Download UnityNeuroSpeech.X.X.X.unitypackage and UNS_StreamingAssets.unitypackage from the official repository. Then you can import them in your project.

They are splitted to avoid importing almost 2GB XTTS model in your project if you only need code/fixes.

Don't forget to put your downloaded Whisper model(.bin) in Assets/StreamingAssets/UnityNeuroSpeech/Whisper/ - yes, it's important.

🎙️ Voice Files

Don’t forget that you need voice files for TTS speech.
Make sure your files meet the following requirements:

  • Format: .wav
  • Duration: 5–15 seconds (longer files work, but TTS will load them more slowly)
  • Contain only one voice and one language, without background noise

Since UnityNeuroSpeech supports multiple voices for multiple agents, files must be named correctly:
<language>_voice<index>.wav

Examples:

  1. English voice, agent index 0 → en_voice0.wav
  2. Russian voice, agent index 3 → ru_voice3.wav

All voices must be placed in:
Assets/StreamingAssets/UnityNeuroSpeech/Voices/

🖼️ Microphone Sprites

You’ll need two sprites for the microphone state (enabled/disabled). But for quick test you can use random default sprites from Unity.