Experimental human like virtual assistant.
A real-time conversational AI system featuring lifelike avatars with synchronized speech, lip movements, and natural interactions.
Virtual Human is an interactive avatar system that combines text-to-speech (TTS), lip-sync animation, and conversational AI to create realistic digital human interactions. Perfect for customer service, virtual assistants, education, entertainment, and accessibility applications.
- Real-time Conversation: Natural language processing for fluid, context-aware dialogue using OpenAI models
- 3D Avatar Rendering: Customizable virtual human characters with realistic expressions using Ready Player Me
- Lip Sync Animation: Accurate mouth movements synchronized with speech audio using SALSA LipSync Suite
- Hybrid Speech-to-Text (STT): First voice it transcribed using lightweight Recognissimo model that also served as a reliable Voice Activity Detection (VAD). Then same audio clip is sent to OpenAI audio transcription API for high quality recogission.
- Text-to-Speech (TTS): High-quality voice synthesis with multiple voice options using RTVoice (with custom Eleven labs provider)
- Full emoji support ❤️
For obvious reasons paid components are not included
- Purchase and install paid components to the default location
- Set API key in Assets\OpenAIKey.cs
- Open Assets\Scenes\Assistant.speech.unity