Trustworthy Voice Interfaces for XR and Human-Centred AI Systems

CandyVoice proposes Trustworthy Voice Interfaces designed for XR, virtual worlds and Human-Centred AI (HCAI), where voice is a primary interaction modality and trust, privacy and safety are critical.

The solution delivers real-time voice interaction layers combining advanced DSP and AI to ensure that spoken interaction with avatars, agents and immersive systems remains reliable, controllable and compliant by design. It addresses key challenges of next-generation XR systems: identity protection, misuse of synthetic voices, robustness in noisy environments and alignment with human values.

The technology includes modular capabilities such as:

• privacy-preserving voice transformation and anonymisation,

• controlled voice rendering for avatars and virtual agents,

• robustness to noise and acoustic variability in immersive settings,

• optional voice integrity and audio deepfake detection,

• low-latency processing suitable for real-time XR interaction.

CandyVoice’s modules are validated on TRL 6–7 demonstrators and designed for edge or local processing, avoiding unnecessary cloud dependency and supporting GDPR and AI Act compliance.

Within Horizon Europe projects, CandyVoice seeks project cooperation to:

• integrate trustworthy voice interfaces into XR platforms, virtual worlds and human-AI interaction frameworks,

• co-develop and validate use cases in training, collaboration, education, accessibility, digital twins or immersive services,

• contribute a specialised audio/voice trust layer to multidisciplinary HCAI and XR consortia.

Target partners include XR platform developers, research organisations in HCI/HCAI, creative industries, system integrators, and Horizon Europe project coordinators focusing on human-centred, trustworthy and inclusive AI.

-------------
Trustworthy AI
Voice interfaces
XR
Virtual worlds
Human-centred AI
HCAI
Real-time voice processing
Privacy-by-design
AI Act
Avatars
Immersive interaction
Horizon Europe