A digital human is a computer-generated person built to look, sound and behave like a real one, usually rendered in 3D with a realistic face and able to speak, listen and react in real time. Most digital humans combine three layers: a visual body (a 3D model with facial animation), a voice (speech recognition and text-to-speech), and a “brain” (scripted dialogue or, increasingly, a large language model). The term covers everything from a bank’s on-screen assistant to a photoreal game character, and it overlaps with, but is not the same as, a virtual influencer or an AI avatar.
At a glance
| Question | Short answer |
|---|---|
| What it is | A lifelike, computer-generated person that can interact, not just appear |
| Core parts | 3D face and body, voice in and out, a dialogue engine |
| Typical uses | Customer service, training, healthcare intake, games, film, brand ambassadors |
| Key tools | Epic’s MetaHuman (now inside Unreal Engine), NVIDIA ACE, studio pipelines |
| Closest cousins | Virtual influencers, VTubers, AI avatars, chatbots |
| Biggest open issue | Disclosure: people should know when the “person” is software |
What does “digital human” actually mean?
The phrase “digital human” is used loosely, so it helps to pin down what most vendors and researchers mean by it. A digital human is defined less by how it is made than by what it does: it presents as a person and it can respond. A still CGI portrait is a render. A pre-recorded animated presenter is a video. A digital human sits in the space where a human-looking face is paired with some ability to take input (your voice, your text, sometimes your facial expression) and answer back.
That definition has two moving parts, and products vary a lot on both:
- Realism. Some digital humans aim for photorealism, down to pores and micro-expressions. Others are stylized but still clearly “human-shaped.”
- Autonomy. At one end, every line is written and approved by people. At the other, a language model improvises answers inside guardrails. Most real deployments sit somewhere in between.
If you want the broader family this term belongs to, our guide to what a virtual influencer is covers the social media side of the same idea.
How does a digital human work?
Under the hood, a modern digital human is a pipeline rather than a single piece of software. A typical real-time setup looks like this:
- Listening. Automatic speech recognition turns what you say into text.
- Thinking. A dialogue engine decides what to say next. This used to be a decision tree; today it is often a large language model connected to a company’s knowledge base.
- Speaking. Text-to-speech turns the reply into a voice.
- Moving. An animation system drives the lips, eyes and face so they match the audio, and adds idle motion so the character does not look frozen between turns.
- Rendering. A real-time engine draws the result on screen, in a browser, an app, a kiosk or a game.
NVIDIA’s ACE microservices are a clear illustration of that stack. When NVIDIA made them generally available in June 2024, the set included Riva for speech recognition, text-to-speech and translation, a Nemotron language model for responses, Audio2Face for facial animation driven by an audio track, and Omniverse RTX for realistic skin and hair rendering. Each piece maps onto one step above.
The hardest part is not any single step but latency. A pause of a second or two between your question and the face reacting is enough to break the illusion, which is why so much engineering effort goes into running these pieces quickly and in parallel.
How are digital humans made?
Creating the visual side of a digital human used to require a film-grade studio, a face-scanning rig and weeks of manual work. That barrier has dropped sharply.
Epic Games’ MetaHuman tools are the best-known example. MetaHuman Creator started as a cloud-based web app that let artists build realistic characters in under an hour. With Unreal Engine 5.6 in 2025, Epic moved MetaHuman creation directly into the engine, and it has since been phasing out the standalone web app, with final shutdown for its last supported engine version scheduled for 5 November 2026. If you built characters in the old web tool, that is the date to export them by.
Beyond the visuals, a production digital human still needs:
- A voice, either licensed from a voice actor, synthesized, or cloned with consent.
- A personality brief: tone, boundaries, what it should never say.
- Content and knowledge: the facts it is allowed to draw on.
- Human oversight: review of transcripts, escalation paths and someone accountable when it gets things wrong.
Real examples of digital humans
Digital humans show up in more places than most people realize. A few well-documented examples help make the category concrete:
- Samsung’s NEON (2020). Samsung’s STAR Labs unveiled NEON at CES 2020 as a project for “artificial humans,” lifelike on-screen characters meant to hold conversations and show emotion. Coverage at the time, such as TechRadar’s, framed NEON as AI companions you might mistake for real people.
- Soul Machines. The New Zealand company, co-founded by double Oscar winner Mark Sagar, built “digital people” for customer-facing roles and was one of the most visible names in the field. In April 2026 it went into receivership owing more than $19.5 million, a reminder that impressive faces do not guarantee a working business.
- Game characters. NVIDIA showcased ACE technology in the game Mecha BREAK, where characters respond to players with generated speech and animation rather than fixed lines.
- Digital models and influencers. CGI personas such as Shudu Gram, often billed as the first digital supermodel, are digital humans in appearance, though most of their output is written and posed by a human team rather than generated live.
Digital human vs virtual influencer vs AI avatar
These terms get used interchangeably, but they describe different things. The simplest way to separate them is by job:
| Term | Main job | Who drives it | Example |
|---|---|---|---|
| Digital human | Interact with people as a lifelike person | Software, often with an LLM, plus human oversight | A bank’s on-screen assistant |
| Virtual influencer | Build an audience and promote things on social media | Usually a human creative team | Lil Miquela, Shudu Gram |
| VTuber | Stream and entertain live | A human performer using motion capture | Kizuna AI |
| AI avatar | Represent a person or brand in video or chat | A real person’s likeness, or a template | A presenter in a generated training video |
| Chatbot | Answer in text or voice | Software | A help-desk bot with no face |
The categories overlap. A virtual influencer can be a digital human if it is given a live conversational layer, and an AI avatar becomes a digital human when it stops reading a script and starts responding. For the streaming side of this map, see our comparison of virtual influencers and VTubers.
Does a digital human need to look realistic?
Realism is the feature that gets the headlines, but the evidence says it matters less than you might expect. The worry behind the push for perfection is the “uncanny valley,” the idea from Masahiro Mori’s 1970 essay that a figure which is almost but not quite human feels eerie, so creators either stay clearly stylized or go all the way to photoreal.
On social platforms, at least, audiences do not seem to punish the in-between. A 2024 study of 33 virtual influencers on Instagram, covering 23,260 posts and published in Humanities and Social Sciences Communications, found no support for the uncanny valley in engagement: humanoid, doll-like, 2D animated and animal characters did not line up in the order the theory predicts. The study’s headline finding was something else entirely: independent virtual influencers out-engaged ones created by brands. How a character communicates, and how transparent it is about what it is, appears to matter more than how lifelike its skin looks.
That is useful news for anyone building a digital human on a budget. A believable voice, quick responses and honest framing may do more for trust than another round of photorealistic rendering.
Where are digital humans used today?
Digital humans have moved out of tech demos and into a handful of practical roles:
- Customer service and sales, as a face on a website or kiosk that can answer common questions.
- Training and simulation, such as practicing a sales call, a medical interview or a difficult conversation with a character that reacts.
- Healthcare and public services, for intake questions and explaining procedures in plain language, with clinical decisions left to people.
- Entertainment, from game characters with dynamic dialogue to digital doubles in film.
- Marketing, through brand ambassadors and digital models.
- Companionship and conversation practice, where AI conversational companion applications, Vinfluencer among them, let people talk with a consistent character in a low-pressure setting.
What are the risks and open questions?
Digital humans raise real questions, and the honest answer is that several are still unsettled.
- Disclosure. The more lifelike the face, the more important it is that people know they are talking to software. Our guide on how to tell if an influencer is AI covers the signals and the platform labels.
- Likeness and consent. Digital humans are often built from scans or voices of real people. Who owns that likeness, and what happens when the contract ends, is an active area of negotiation in film, games and advertising.
- Accuracy. A friendly face can make wrong answers more persuasive. Any digital human connected to a language model needs limits on what it will claim.
- Emotional weight. People respond to faces. That can make a digital human more helpful, and it can also make attachment stronger than a text bot would create, which is a design responsibility, not just a feature.
- Durability. The field has seen high-profile companies stumble, and tools change quickly. Anyone investing in a digital human should plan for exporting their assets.
FAQ
What is a digital human in simple terms?
A digital human is a computer-generated person that looks and sounds human and can respond to you, usually by listening to your voice or text and answering out loud with a moving face. Think of it as a chatbot with a realistic body, voice and expressions.
Is a digital human the same as an AI?
Not exactly. A digital human is the whole character: face, voice, animation and dialogue. AI is often one part of it, typically the speech recognition and the language model that writes replies. Some digital humans use little AI and run mostly on scripts written by people.
What is the difference between a digital human and a virtual influencer?
A digital human is built to interact with people in real time, while a virtual influencer is built to grow an audience on social media. Many virtual influencers look like digital humans, but most of their posts are created by a human team rather than generated in live conversation.
How do you create a digital human?
Most creators start with a 3D character tool such as Epic’s MetaHuman, now built into Unreal Engine 5.6 and later, then add a voice, a dialogue engine such as a large language model, and facial animation driven by audio. A personality brief and human review are what make it usable.
Are digital humans safe to talk to?
Digital humans are generally as safe as the software and policies behind them. Check who runs it, what it does with your data, and whether it clearly says it is not a person. Treat its answers on health, money or legal topics as a starting point and confirm them with a qualified human.
Discover more from Vinfluencer
Subscribe to get the latest posts sent to your email.