Neuphonic & Rapport launch real-time digital humans

Neuphonic has partnered with Rapport to launch a real-time digital human system that runs on standard CPU hardware. It is designed to deliver photorealistic avatars with no perceptible delay.

The partnership combines Neuphonic’s text-to-speech software with Rapport’s facial and full-body animation technology. Together, the system can generate voice, lip movement and facial performance in under 100 milliseconds while running fully on-device, rather than through cloud services or specialist graphics processors.

The approach addresses a long-standing hurdle for interactive digital characters, where slow response times and reliance on expensive hardware have limited wider adoption. By using standard CPUs, the companies are targeting devices and environments where dedicated GPU infrastructure is too costly or impractical.

Rapport is a division of Speech Graphics, whose animation tools have been used in games including Call of Duty, The Last of Us Part II and Hogwarts Legacy. London-based Neuphonic focuses on voice AI models built to run on consumer devices.

The system could be used in games, film and virtual production, advertising, training and customer engagement. The companies also pointed to education and virtual assistants as areas where more responsive digital characters could be deployed at greater scale.

On-Device focus

A central part of the tie-up is that processing remains local to the device. This means user data does not need to leave the immediate environment for cloud processing, reducing latency and lowering infrastructure requirements.

Keeping the system on-device also reflects a broader shift in AI development, as companies seek to balance faster response times with privacy concerns and rising cloud costs. In sectors such as gaming and interactive media, real-time responses have become increasingly important as users expect conversations and character reactions without obvious lag.

Neuphonic said its speech technology runs with sub-100 millisecond latency on CPUs. Rapport contributes real-time facial and body animation designed to match speech output closely enough to create natural-looking interactions.

“We’re making real-time digital humans practical,” said Sohaib Ahmad, Chief Executive Officer and Co-founder of Neuphonic. “For the first time, you can run a fully interactive, lifelike avatar instantly on everyday hardware. It shifts this from something that required high-end infrastructure to something that can run anywhere, and will unlock entirely new use cases across games, media and AI. It will move us closer to AI that feels like a natural, seamless extension of human interaction.”

Speech Graphics has spent more than a decade developing audio-driven animation for entertainment, with work ranging from pre-produced character performances to conversational systems. The partnership with Neuphonic shows how that expertise is moving into live interactive settings rather than remaining confined to traditional game and film pipelines.

Wider uses

For game developers, the technology could support non-player characters that speak and respond dynamically during play. In film and virtual production, it could enable live digital performances without remote processing. In advertising and customer engagement, it could support interactive spokespeople or branded characters.

The ability to deploy across ordinary hardware could also make digital humans easier to distribute at scale. The companies said the system could run on billions of devices, widening access beyond studios or organisations with substantial graphics budgets.

The partnership has received support from Innovate UK’s Creative Catalyst programme, which backs projects in the creative and technology sectors. That support underlines growing interest in digital human systems that can operate in real time and in practical commercial environments.

“Neuphonic was a natural partner because the technologies are highly complementary,” said Gregor Hofer, Chief Executive Officer and Co-founder of Rapport, a division of Speech Graphics Ltd. “They bring ultra-low latency, on-device voice that runs efficiently on CPU, and we bring production-proven, real-time facial animation. Together, this enables fully local digital humans that can speak, animate and respond instantly, without relying on the cloud, opening up new possibilities for performance, cost and deployment at scale. It’s a complete, real-time conversational experience.”