At the BMW Group’s battery assembly plants, virtual reality is helping employees prepare for the production of sixth-generation high-voltage batteries. In a detailed digital replica of the factory, production staff can practice real manufacturing operations under realistic conditions before working with the actual equipment for the first time. An AI-controlled avatar guides them through the Gen6 battery assembly process, explains individual production steps and answers questions in natural conversation – making virtual training more interactive, practical and accessible.
The AI agent is a dynamic, generative AI-based NPC (Non-Plyable Character) that converses with users in a natural and context-sensitive manner. Whether it’s explaining the complexity of battery assembly, answering questions about business processes or handling several conversations at once, the AI agent is a key element of the virtual environment, where people and AI work together seamlessly.
The challenge: How can AI communicate naturally?
To ensure that the AI agent can communicate with staff as naturally as possible, it must integrate seamlessly into the voice chat environment. It must be easy to add to calls, and the conversation itself must feel as natural as possible.
The key factors are the AI’s voice and its ability to interact. In an immersive 3D environment, it is not enough for all voices to resonate equally from the speaker, as in a conventional Teams call. Instead, the AI NPC must be spatially audible. If it is standing on the left, its voice should come from the left. When it moves, the volume and direction of its voice must adjust accordingly. And when 200 users are in the environment at the same time, communication must work without lag or any loss of quality.
So, the question for companies is: How do you integrate an AI into a 3D environment that not only provides answers, but also sounds natural, is spatially perceptible and can interact with any number of users? The solution needs to be scalable, stable and easy to integrate.
After all, if the AI NPC sounds like a robot or the communication comes across as unnatural, the entire 3D environment loses its added value. Users should be able to concentrate fully on the content without being distracted by technical shortcomings.
The solution: ODIN Voice as an enabler for real interaction
The answer to these challenges is ODIN Voice. With this technology, the AI NPC becomes a fully-fledged member of the 3D environment:
- Spatial & High-Quality Voice: Voices come from the speaker’s direction—whether AI NPC or colleague. Movements are dynamically rendered in real time. 48 kHz and Opus Codec ensure clear, lifelike voices
- Self-Hosted & Data Sovereign: ODIN Voice can be operated on-premise within your own infrastructure—GDPR-compliant and free from third-party dependencies.
- Cross-Platform: With ODIN Voice you have an SDK for all major platforms and engines, ensuring plug-and-play integration into existing systems.
- Scalability: ODIN Voice works with one, 200 or even over 1,000+ users – without any loss of performance.
From Code to Colleague: How the BMW Group Brings AI NPCs to Life
Thus, ODIN Voice is part of a use case involving AI-powered NPCs within the BMW Group’s 3D environment. The example illustrates how immersive voice communication can support interaction between users and AI-driven avatars.
Potential use cases include guided experiences, training scenarios and collaborative review sessions in virtual environments. In these scenarios AI NPCs feel like real colleagues and users can focus on the content, the discussion, and the creativity—not the technology behind it. ODIN Voice is the technology that makes this possible, powering natural, immersive communication between humans and AI in 3D environments.
Want to make virtual feel real?
Voice chat that connects teams naturally. Hosting that scales globally. Smart solutions for transcription, meeting minutes and moderation.