By Sagar Shankaran, Founder of CallSphere
NVIDIA launches ACE (Avatar Cloud Engine) microservices, allowing enterprises to deploy photorealistic AI agent avatars with real-time speech, emotion, and gesture capabilities.
Key takeaways
NVIDIA has launched ACE (Avatar Cloud Engine) Microservices for general availability, a suite of cloud-native APIs that enable enterprises to deploy photorealistic AI agent avatars with real-time speech synthesis, facial animation, emotional expression, and gesture generation. The platform, announced at GTC 2026 on March 11, transforms how businesses create interactive AI experiences by providing the visual and conversational layer that turns text-based AI agents into lifelike digital humans.
ACE has been in development since 2023, with early previews demonstrating digital human capabilities for gaming and entertainment applications. The microservices release marks a strategic pivot toward enterprise use cases, with NVIDIA positioning ACE as the standard infrastructure for AI-powered customer interactions across healthcare, financial services, retail, hospitality, and education.
NVIDIA ACE Microservices is composed of six core services that work together to create a complete digital human experience:
flowchart LR
INPUT(["User intent"])
PARSE["Parse plus<br/>classify"]
PLAN["Plan and tool<br/>selection"]
AGENT["Agent loop<br/>LLM plus tools"]
GUARD{"Guardrails<br/>and policy"}
EXEC["Execute and<br/>verify result"]
OBS[("Trace and metrics")]
OUT(["Outcome plus<br/>next action"])
INPUT --> PARSE --> PLAN --> AGENT --> GUARD
GUARD -->|Pass| EXEC --> OUT
GUARD -->|Fail| AGENT
AGENT --> OBS
style AGENT fill:#4f46e5,stroke:#4338ca,color:#fff
style GUARD fill:#f59e0b,stroke:#d97706,color:#1f2937
style OBS fill:#ede9fe,stroke:#7c3aed,color:#1e1b4b
style OUT fill:#059669,stroke:#047857,color:#fff
This service takes streaming audio input — either from a text-to-speech engine or a human voice — and generates photorealistic facial animations in real time. The system maps audio features to over 250 individual facial muscle movements (blendshapes), producing animations that accurately reflect speech patterns, emotional tone, and natural micro-expressions.
The latest version supports 40 languages and can generate facial animations with less than 80 milliseconds of latency, enabling natural conversational interactions without perceptible delay. NVIDIA claims this represents a 5x improvement over the previous generation and approaches the threshold of human-imperceptible latency.
NVIDIA's Riva platform provides both automatic speech recognition (ASR) and text-to-speech (TTS) capabilities. The TTS component generates natural-sounding speech from text with controllable parameters including speaking rate, pitch, emphasis, and emotional tone. Riva supports voice cloning, allowing enterprises to create custom brand voices from as little as 30 minutes of reference audio.
For the ASR component, Riva processes incoming user speech with streaming transcription, enabling real-time conversational interactions. The system handles overlapping speech, background noise, and accented English with 97% accuracy — on par with or exceeding human transcriptionist performance.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
ACE Microservices integrate natively with NVIDIA's Nemotron family of language models, which power the conversational intelligence behind digital human agents. Nemotron models are optimized for low-latency inference on NVIDIA GPUs, enabling response generation in under 200 milliseconds for typical conversational turns.
The integration also supports third-party LLMs including models from OpenAI, Anthropic, Google, and open-source alternatives, providing flexibility for enterprises with existing AI investments.
Tokkio is the orchestration layer that manages the complete interaction flow between a user and a digital human agent. It handles turn-taking (knowing when the user has finished speaking), manages conversation state, triggers appropriate emotional responses based on conversation context, and coordinates the various microservices to maintain a coherent, natural interaction.
Tokkio supports both one-on-one interactions and group scenarios where a digital human agent interacts with multiple users simultaneously — useful for kiosk deployments, virtual receptionist scenarios, and digital classroom environments.
NVIDIA Maxine provides video processing capabilities including background replacement, lighting normalization, eye contact correction, and super-resolution. For ACE deployments, Maxine ensures that digital human agents appear consistently across different display devices and environments, from mobile phones to large interactive displays.
This service manages the 3D avatar assets, including character models, clothing, environments, and animation libraries. Enterprises can choose from a catalog of pre-built avatar designs or create custom characters using NVIDIA Omniverse tools. The service supports both realistic human avatars and stylized character designs.
Several high-profile enterprise deployments are already live:
A major US hospital network has deployed ACE-powered digital human agents at emergency department check-in kiosks. The avatar conducts initial patient intake interviews, collects symptom information, assesses urgency using clinical triage protocols, and provides wait time estimates. The system supports 12 languages and is specifically trained to communicate with patients who may be anxious, in pain, or confused.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
"The digital human agent handles 60% of our intake volume during peak hours," reported the hospital network's Chief Digital Officer. "Patient satisfaction scores for the AI intake experience are actually 8 points higher than human intake, primarily because wait times are eliminated and the interaction is private."
A global bank has integrated ACE avatars into its mobile banking application, providing a digital financial advisor that can discuss portfolio performance, explain market conditions, and walk customers through complex financial products. The avatar maintains a consistent personality and remembers previous conversations, creating a relationship-like dynamic that the bank reports has increased customer engagement with advisory services by 156%.
Multiple luxury retail brands have deployed ACE digital humans in flagship stores, where interactive displays feature lifelike AI assistants that can discuss product details, recommend complementary items, check inventory, and process orders. The avatars are designed to embody the brand's aesthetic and communication style, providing a premium experience that extends the brand's identity into the digital realm.
An online education platform has created subject-specific digital human tutors that conduct one-on-one tutoring sessions. Each tutor avatar has a distinct personality, teaching style, and area of expertise. The platform reports that students who interact with avatar tutors complete 40% more course material and score 18% higher on assessments compared to text-only AI tutoring.
ACE Microservices run on NVIDIA's cloud infrastructure or can be deployed on-premises using NVIDIA DGX or certified server hardware. The minimum configuration for a production deployment requires an A100 or H100 GPU, with each GPU supporting approximately 16 concurrent avatar sessions.
Pricing follows a consumption model:
"We deliberately chose per-minute pricing to make adoption frictionless," said Rev Lebaredian, VP of Omniverse and Simulation at NVIDIA. "A company can start with a single kiosk pilot and scale to thousands of endpoints without renegotiating contracts."
NVIDIA's entry into the digital human market puts pressure on existing players including Soul Machines, UneeQ, and Synthesia, which have offered AI avatar platforms for several years. While these companies have established customer bases and proven technology, NVIDIA's advantages in GPU-accelerated inference, end-to-end stack integration, and brand recognition in the enterprise AI market represent a formidable competitive challenge.
"NVIDIA is not just entering the digital human market — they are defining the infrastructure layer that everyone else will build on," said Matthew Ball, CEO of Epyllion and author of "The Metaverse." "This is similar to what NVIDIA did with CUDA for GPU computing. They are creating the standard."
Written by
Sagar Shankaran· Founder, CallSphere
Sagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
A2A unlocks cross-vendor agent coordination, but most enterprise voice/chat workloads still ship faster on a single-vendor stack. Here is how to choose.
Working memory, permanent memory, sandboxes, harnesses, governance — the practical blueprint enterprises are using to ship long-horizon AI agents in 2026.
Anthropic confirmed JPMorgan Chase, Goldman Sachs, Citi, AIG, and Visa in production on Claude as of May 2026. What each pattern of usage looks like.
At Cloud Next 2026 Google renamed Vertex AI to Gemini Enterprise Agent Platform and absorbed Agentspace. What actually changed and why a rebrand made sense.
AI Control Tower is the governance layer for ServiceNow's Project Arc — policy, monitoring, and audit logs for autonomous agents. Here is how it works.
CAISI announced new agreements with Google DeepMind, Microsoft, and xAI in May 2026. What gets tested, what changes for enterprise AI buyers, what to watch.
© 2026 CallSphere LLC. All rights reserved.
Made within New York
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI