Avatar

Template:About An avatar is the digital representation of a user inside a virtual space. In Virtual reality and augmented reality it is the body, or part of a body, that other people see in place of the user and that the user looks down at and controls as their own. The avatar is what turns an abstract account into a visible presence: it gives a person a location in the scene, a way to gesture and make eye contact, and an identity that other participants can recognise and address. Avatars range from a pair of disembodied floating hands to a fully tracked, photoreal human figure, but in every case the goal is the same, to stand in for the user so that interaction in the virtual world feels like interaction between people.

Avatars are central to Social VR and to the broader idea of the Metaverse, where the point of the system is for users to meet, talk, and do things together in a shared three-dimensional space rather than to look at a flat screen.

Origin of the term

The word avatar comes from the Sanskrit avatara, which in Hinduism refers to the descent of a deity into a terrestrial or bodily form. The computing sense, an on-screen figure that stands in for a real person, emerged in the 1980s. The game designer Richard Garriott is generally credited with the first use of the word in this sense, in the 1985 role-playing game Ultima IV: Quest of the Avatar; he chose the Hindu word because he wanted the player's on-screen character to be the player's own self made manifest in the virtual world rather than a separate fictional persona.

The term took on its modern networked meaning in Habitat, an early online graphical role-playing world created by Chip Morningstar and Randy Farmer at Lucasfilm Games and launched in 1986. Habitat used "avatar" to mean the animated figure that represented a connected player in a shared online space, which is very close to how the word is used today.

The word entered popular culture through Neal Stephenson's 1992 cyberpunk novel Snow Crash, which used "avatar" for the virtual bodies that people inhabit inside its fictional Metaverse. Stephenson noted in the book's acknowledgments that the term was already in use in the Habitat virtual-reality system, so he did not coin it; his novel popularised it. Snow Crash is widely cited as the source of both the words "avatar" and "metaverse" as they are now used in the VR industry.

Avatar forms in VR

Because consumer VR hardware tracks only a few points on the body by default, namely the head and the two hands holding the controllers, avatars in VR fall into a small number of forms that reflect how much of the user can actually be sensed.

Form	What is shown	Typical use
Hands only	A pair of floating hands or controllers, with no arms or torso	Minimal experiences and tools where only pointing and grabbing matter
Head and hands	A floating head and two floating hands, with the space between left empty	Many early social apps; cheap to drive from a standard headset and two controllers
Upper body	Head, torso, and arms, usually with the lower body omitted or faded out	The default for several mainstream social platforms; the arms are reconstructed by software
Full body	A complete figure including legs	Requires either extra tracking hardware or software that estimates the unseen joints

The "floating" head-and-hands avatar became common because it can be driven directly from the three points a basic VR rig provides without guessing anything. It also sidesteps a problem that troubled early full-body avatars: when software has to invent the position of the elbows, hips, and legs and gets it wrong, the result looks unnatural and breaks the user's sense of having a real body. For years the major social platforms deliberately left the legs off their avatars for this reason, a design choice that drew frequent ridicule before tracking and estimation improved enough to add them back.

Driving avatars from tracking

What makes a VR avatar feel alive, rather than like a puppet, is that it moves in real time with the user. The headset and controllers report their position and orientation continuously, and the avatar's head and hands are mapped directly onto that data, so when the user turns their head or waves, the avatar does the same with very little delay. The unmeasured parts of the body have to be filled in.

Reconstructing the body

With only the head and two hands tracked, software estimates where the rest of the limbs must be using inverse kinematics, which works backwards from the known positions of the hands and head to compute plausible positions for the elbows, shoulders, and spine given the lengths of the bones and the limits of how human joints bend. This three-point approach can drive an upper body convincingly, but it cannot know what the legs are doing, which is why estimated legs were historically unreliable. Adding dedicated trackers worn on the waist and feet gives the system real data for those joints and allows a faithful full body. In VRChat, for example, the headset tracks the head and the controllers track the hands, so a common setup adds three more trackers on the hips and feet for six-point full-body tracking, with support for more trackers for higher fidelity. More recent research estimates a full-body pose from just the head and hands using machine learning trained on motion data, reducing the need for extra hardware.

Face, eyes, and hands

Higher-end headsets add sensors that let the avatar carry the user's face as well as their movement. The Meta Quest Pro, introduced in 2022, uses inward-facing infrared sensors, three aimed at the eyes and upper face and two at the lower face, and a trained model reads them to drive an avatar's gaze and expression, so it can blink, smile, and make eye contact in step with the user. For privacy the system does not hand applications a picture of the face; it gives them a set of values describing generic movements, such as how much the nose is scrunched or the brows are raised, which the application maps onto its own avatar, whether cartoonish or realistic. Eye tracking also lets the avatar's eyes look where the user is looking, which improves the perceived quality of conversation in VR, and Hand tracking using the headset cameras lets the avatar's fingers move individually without controllers. Tracking the face this way matters socially because so much human communication is carried by gaze and expression, and an avatar that reproduces them supports the feeling of actually being with another person.

Embodiment and presence

A well-driven avatar does more than show the user to others; it can make the user feel that the virtual body is genuinely their own. This is studied as the sense of embodiment, defined by Kilteni and colleagues in 2012 as the set of sensations that arise from being inside, having, and controlling a body, and broken into three parts: the sense of self-location (feeling positioned inside the body), the sense of agency (feeling that you cause its movements), and the sense of body ownership (feeling that the body is yours).

The roots of this work lie in the rubber hand illusion reported by Botvinick and Cohen in 1998, in which a person whose real hand is hidden comes to feel that a visible rubber hand is their own when both are stroked at the same time; the illusion appears only when the stroking is synchronous. In VR the same principle extends to the whole body: seeing a life-sized virtual body from a first-person viewpoint, moving in time with you, is generally enough to make it feel like your body. A strong sense of embodiment in turn reinforces presence, the feeling of really being in the virtual place, because a believable body in the scene strengthens the sense of being there. The VR & AR Wiki article on Presence covers the embodiment and body-ownership research in more detail, including experiments showing that the kind of avatar a person is given can change their behaviour and attitudes.

Major avatar systems

Several avatar systems are widely used across consumer VR.

System	Platform	Notes
VRChat avatars	VRChat (PC and standalone)	Avatars are built by users in a 3D tool and uploaded through the platform's SDK; VRChat states that its worlds and the millions of avatars in them are all created by users. Avatars can take almost any form, from a human to "an alien" or "a talking dog," and flow with the user's movement, with support for full-body and finger tracking.
Meta Avatars	Meta Horizon platform and third-party apps	Meta's cartoon-styled avatars, used in Horizon Worlds and offered to other developers through the Meta Avatars SDK. Legs were added to the system in 2023 after the upper-body-only avatars were widely mocked.
Ready Player Me	Cross-app (Unity, Unreal, web)	A cross-game avatar service that lets a user build one avatar, optionally generated from a single photo, and reuse it across many separate apps and games rather than rebuilding it in each.
Apple Personas	Apple Vision Pro	Apple's realistic, scanned representation of the user's own face, used in FaceTime and other apps on Vision Pro.

VRChat

VRChat is the largest user-generated social VR platform, and its model is built around custom avatars. Rather than choosing from a fixed set, users create an avatar in a 3D modelling tool and upload it through VRChat's software development kit, which means the avatar can be literally anything the creator can model. The platform describes its content as "hundreds of thousands of worlds, millions of avatars, all created by the users," and frames the avatar as a way to "be anything" and explore your identity. VRChat avatars can be driven by full-body tracking and per-finger hand tracking, and face tracking can be routed into an avatar's expressions through Open Sound Control on supported hardware, so a properly prepared avatar mirrors the user's face as well as their body.

Meta Avatars and Horizon

Meta Avatars are the stylised, cartoon-look avatars used across Meta's social products, including Horizon Worlds, and made available to outside developers through the Meta Avatars SDK. For years they showed only the upper body, and the legless figures became a running joke about the company's metaverse effort. Meta added legs to the avatars in 2023, first in the Quest home environment and then in Horizon Worlds, and exposed the capability to third-party apps through an updated version of the SDK. In 2024 Meta overhauled the avatar system further with more customisation, including adjustable body proportions and facial features.

Separately, Meta's Reality Labs Research runs a long-term research project called Codec Avatars, which aims to produce photorealistic, lifelike avatars for telepresence rather than the cartoon style shipped in current products. The two are distinct: the cartoon Meta Avatars are what people use today, while Codec Avatars are an experimental effort toward avatars that look indistinguishable from a real person.

Ready Player Me

Ready Player Me is a cross-game avatar service whose purpose is to let a single avatar follow a user between unrelated applications. A user creates one avatar, which can be generated automatically from a single photograph and then customised, and that avatar can be used across a large number of apps and games rather than being rebuilt for each one. Developers integrate it through software development kits for Unity, Unreal Engine, and the web, and Ready Player Me states that its avatars are supported in tens of thousands of apps and games. The model addresses a fragmentation problem in social VR, where each platform historically had its own incompatible avatar.

Apple Personas

On the Apple Vision Pro, the user's avatar is called a Persona. Unlike the systems above, a Persona is meant to look like the actual user: it is produced by holding the headset out at arm's length and letting its front sensors scan the face while the user turns their head and runs through a set of guided expressions, after which Apple's software builds a digital likeness. Apple describes a Persona as "a digital representation of themselves created using Apple's most advanced machine learning techniques, which reflects face and hand movements in real time." In a FaceTime call, other participants see the Persona instead of a camera feed. Apple later added spatial Personas, which place the caller's three-dimensional Persona in the room rather than inside a flat window, for up to five participants at once.

Customisation and identity

Beyond their technical role, avatars are how people express who they want to be in a virtual world. Most systems let users adjust appearance in detail, from body shape and facial features to hair and clothing, and platforms built on user-generated content go further by letting people import or commission entirely custom avatars. The avatar a person chooses is often not a faithful copy of their physical self; many users deliberately present as a different age, a different look, a fantasy creature, or a different gender, and the freedom to do so is part of the appeal of Social VR. This is a point of tension with the realistic-scan approach taken by Apple Personas, which aims for accurate self-representation, and it is why some platforms offer both stylised and realistic options for different situations. Because an avatar is the thing other users see and remember, it functions as a persistent online identity, and on cross-app services like Ready Player Me a single avatar can carry that identity across many separate virtual worlds.

See also

• Social VR
• Metaverse
• Presence
• VRChat
• Apple Vision Pro
• Hand tracking
• Eye tracking
• Virtual reality