Unity VR AI Voice Assistant
Budget: €30 – €250 EUR
I’m building SpeakVR, a Unity-based VR experience that lets users chat naturally with an AI companion. The focus is casual conversation, so the assistant must answer general questions, engage in small talk, correct a user’s grammar when needed, and slip into light role-play (for instance, pretending to be a customer in a practice scenario).
My foundation is in place: a simple VR scene, OpenAI keys and usage rights, plus the intention to rely on Whisper for speech recognition, ChatGPT for dialogue, and a lifelike text-to-speech engine for the final voice reply. What’s missing is the glue that turns these pieces into a seamless, low-latency conversational loop.
Here’s what I need from you:
• C# scripts that capture microphone input inside Unity, stream it to Whisper, and return clean transcriptions
• Prompt-building logic that feeds context-aware text to ChatGPT and receives a coherent reply that can include gentle grammar corrections
• Integration of a high-quality TTS service, with spatialised playback in the VR scene so the voice feels present
• Simple inspector fields for API keys, temperature, and prompt tweaks, along with a toggle for on-screen captions
• A small demo scene showing everything working end to end, plus clear comments and a short README so I can extend the project myself
Efficiency matters; the interaction needs to feel fluid, not laggy. If you know your way around asynchronous calls (UniTask, coroutines, or similar) and are comfortable optimising network traffic, let’s talk.
My foundation is in place: a simple VR scene, OpenAI keys and usage rights, plus the intention to rely on Whisper for speech recognition, ChatGPT for dialogue, and a lifelike text-to-speech engine for the final voice reply. What’s missing is the glue that turns these pieces into a seamless, low-latency conversational loop.
Here’s what I need from you:
• C# scripts that capture microphone input inside Unity, stream it to Whisper, and return clean transcriptions
• Prompt-building logic that feeds context-aware text to ChatGPT and receives a coherent reply that can include gentle grammar corrections
• Integration of a high-quality TTS service, with spatialised playback in the VR scene so the voice feels present
• Simple inspector fields for API keys, temperature, and prompt tweaks, along with a toggle for on-screen captions
• A small demo scene showing everything working end to end, plus clear comments and a short README so I can extend the project myself
Efficiency matters; the interaction needs to feel fluid, not laggy. If you know your way around asynchronous calls (UniTask, coroutines, or similar) and are comfortable optimising network traffic, let’s talk.