Nvidia opens early access to its digital human cloud engine

Bringing a virtual character to life usually takes weeks of painstaking work: detailed 3D rigging, motion capture actors, and hours in an audio studio syncing mouth shapes to every phoneme.

NVIDIA wants to turn that entire pipeline into a couple of cloud API calls. The company just launched early access for its Omniverse Avatar Cloud Engine, or ACE for short.

First teased back in September with a handful of close partners, ACE is now opening up to developers who want to build and deploy intelligent, interactive avatars at scale.

Lip-syncing on the fly from raw audio

At its core, ACE is a collection of cloud-native AI microservices built on top of NVIDIA Unified Compute Framework. Instead of building your own avatar tech from scratch, it gives you the modular building blocks to inject intelligence and realistic animation into almost any digital character.

The standout service is Audio2Face. You feed it a standard audio file, and the generative AI automatically creates matching, realistic 3D facial animation. It hooks directly into rendering engines like Epic Games' Unreal Engine without manual keyframing.

For 2D work, a microservice called Live Portrait lets creators drive 2D illustrations or stylized faces using a live camera feed. Meanwhile, speech generation is handled by NVIDIA Riva TTS, which turns plain text transcripts into natural-sounding voiceovers without needing manual cues for inflection or timing.

Because it runs as cloud microservices, developers can stream these interactive avatars into pretty much any endpoint. That includes VR headsets, tablets, or self-service kiosks at quick-service restaurants.

Ready Player Me and MetaHumans tap in

NVIDIA is already showing off integrations with established 3D platforms. Cross-game avatar maker Ready Player Me is using the animation microservice to bring its characters to life.

Ready Player Me CEO Timmu Tõke noted that after spending seven years building an avatar system for apps and games, ACE makes it dramatically easier for developers to animate those characters automatically.

Epic Games is also leaning in with MetaHuman. Pairing MetaHuman's photorealistic facial rigs with NVIDIA's automated cloud pipeline drastically cuts the time needed to create conversational virtual humans. Even VTubers are being targeted, with tools to upgrade flat 2D streaming avatars into fully animated, cloud-rendered 3D characters.

Why it matters

Building believable digital humans has always been a painful trade-off: you either spend a fortune on high-end rigging, or you end up with lifeless, uncanny marionettes. We think NVIDIA's move here is brilliant because it is not trying to sell yet another character creator. Instead, it is positioning itself as the unseen AI engine underneath the tools developers already use, from Unreal Engine to Ready Player Me.

For indie studios, retail brands, and interactive creators, this lowers the bar to putting responsive 3D assistants on screens anywhere. But there is a catch: relying entirely on cloud-hosted microservices means your interactive characters live and die by latency and connection stability. Running millions of real-time Audio2Face and TTS streams across thousands of kiosk screens could rack up steep cloud bills.

Strategically, NVIDIA is making sure that as conversational AI moves from text boxes to 3D embodiments, its hardware and software remain the default backbone. By solving the messy animation and speech integration problems in the cloud, NVIDIA is turning what used to be a specialized VFX pipeline into everyday infrastructure.