Discover more from Building Virtual Worlds
"This is Me": Avatars and the Future of Digital Identities
Generative AI is bridging the gap between games and social networks
In November 2022, AI-generated portraits suddenly flooded across the internet, created using a 4-year old photo editing app call Lensa. These whimsical photos were enabled with generative AI1, and allowed users to create high quality digital art versions of themselves in minutes (and for $7.99). Two thoughts went off in my head, the first was that my local central park caricature artist was soon to be out of a job, and the second was that these were not just photos - these were rudimentary avatars, avatars that I have seen my whole life within virtual worlds.
What Are Avatars and Why Should We Care?
The word avatar derives from the Sanskrit word for “descent”, and refers to the descent of a deity to the human realm. In today’s world of technology, avatar generally refers to the digital representation of a user (the self) in a digital enviornment, whether that is a 3D video game world or the broader internet.
However, avatars can be more than just identity markers. Analogous to the mouse cursor in a PC operating system, the avatar is also a critical enabler of human-computer interaction - they are what enable us to interact with and change digital worlds.
Within digital worlds2, avatars often function as both an extension of your controller and as your digital identity: how users view themselves and project their identity to others. These two use cases can be one or the other or increasingly, both:
To lifelong gamers, the blurring of identity between the physical and virtual world should come as no surprise. The surprise is that generative AI and other personalization technology are driving up avatar adoption rate, particularly among people who did not previously consider themselves denizens of virtual worlds. In this way, avatars may be a gateway drug, of sorts.
Psychology of Avatars
Over the years, there has been numerous studies in the psychology behind avatars, but the major effects to note is the Proteus Effect – the behavior of people in virtual worlds is changed by characteristics of their avatar3. Hence an avatar not only impacts the digital world but also impacts the action of its user as well (a feedback loop). Users with avatars that are more similar to their real-world self also report a greater sense of immersion in the digital world.
In real life, the presence of one’s physical body in social encounters prevents people from claiming identities that are inconsistent with their physical attributes or is inconsistent with what others already know about them.
In the digital world, thanks to disembodiment and pseudo-anonymity, virtual identities can be much more flexible and expressive - instead of a digital copy, you can be a “possible self” instead. Oftentimes, this results in embellishments in one’s existing talents or physical appearance, but can also include changing the job/role you play in life (GTA V Roleplay). You become who you want to be, as least for the moment.
This explains why Lensa and other AI-powered avatar generators provides users a dopamine hit - it’s uniquely based on how your existing appearance, but renders a fanciful version of you, allowing you to imagine the version of “you” who would look like that.
In summary, personalized avatars can help:
Create a greater sense of self-presence and immersion in virtual worlds (“this is me, and I am here”)
Make user happier about themselves (“this is who I should be”)
Stimulate creativity (“I am no longer bound by the limitations of my real-world identity”)
Avatars is already a major business spanning multiple industries. In 2020, the avatar industry was worth ~$50B, of which ~$40B is in gaming (this is about 20% of the entire gaming industry’s sales!)4, primarily comprised of in-game cosmetic sales.
We expect the avatar market to outgrow the rest of gaming, especially as games transition to a GaaS (games-as-a-service) model. We are also excited about emerging use cases for avatars outside of “core” gaming.
History and Use Cases of Avatars
The primary use cases of avatars today is in gaming and social media, although there are emerging use cases in both consumer (e.g., brand marketing, music) and B2B (e.g. chatbots for customer service, healthcare, etc.).
The concept of an avatar has been a part of games since the very early days of the industry, starting with the “controller” function. One can argue that the player-controlled cannon in 1978’s Space Invader was an avatar. However, the term was only applied to the modern “identity” function later, with 1985’s Ultima IV: Quest of the Avatar, an early role-playing game (RPG).
Today, gaming avatars have grown in realism, but still take two forms (and functions)
Non-customizable avatars for simulation (FIFA, Red Dead Redemption 2) - prioritize the “controller” function
Highly customizable avatars for role-playing (World of Warcraft, VRChat) - prioritize the “identity” function
While some games do not need avatar customization (generally pure simulation games or narrative-driven games), the general trends is towards greater customization and personalization - towards the “identity” function. This trend is due largely in part to gaming’s transition to become both more social and service-oriented.
Roblox5 is a good example of the evolution of gaming avatars:
It’s hard to imagine in today’s personalized and user data-rich internet, but the early days of the internet (internet 1.0, pre-Google, Facebook) was wild6, and users preferred to remain anonymous or at best, pseudonymous. In this world, avatars took the form of 2D icons within early internet forums (Slashdot, Newgrounds) and instant messenger applications; in other words, profile picture. In a pseudonymous environment, the avatar was the only way to present your personality to strangers on the internet7. This use case is still relevant today (think the PFT NFT craze of 2021).
When Myspace and modern social media ushered in the era of real-name identities in the internet, avatars took a back seat to profile photos, particularly in the west8. However, in the era of visual-focused social networks (Snap, Instagram, TikTok), we have seen a trend towards less realism across social networks, from filters to masks and more recently, full-on personalized avatars such as Bitmoji, TikTok Avatars, etc.9.
On the more extreme end of the trend, a growing number of live streamers called "vTubers" (short for virtual YouTuber) have fully embraced a new digital-only identity with the aid of motion capture technology. The number of vTubers reached 16,000 as of late 202110, with top vTuber Gawr Gura having over 4M subscribers. San Francisco-based “VTubing” agency Vshojo raised $11 million in 2022.11
Music and Transmedia
Avatars have enabled creators to extend their influence into digital worlds and to experiment with alternative new identities within digital worlds.
The music industry is no stranger to this, with examples like Gorillaz and K/DA. Gorillaz (around since 1998!) has been highly influential, selling millions of records and winning multiple awards including a Grammy12. K/DA is a virtual K-pop band formed in 2018 with characters from the video game League of Legends but voiced by popular artists including Madison Beer and members of the K-pop group (G)I-dle. The popularity of K/DA has boosted the cultural relevance of League of Legends (in addition to generating in-game cosmetic sales), and illustrates a successful case of a video game IP branching into a new media format.
Avatars have also blurred the line between real-life and virtual representations of artists. We’ve seen many popular musicians including Travis Scott and Ariana Grande hosting virtual concerts in virtual worlds such as Fortnite and Roblox. Travis Scott’s 2020 Fortnite concert reportedly grossed $20M including digital cosmetic sales, compared to $1.7M for a single live concert, and reached nearly 28M unique viewers (or are they players?).
Emerging Use Cases
Within B2B, digital avatars are starting to gain adoption as digital brand ambassadors. Lu, the virtual face of Brazilian retailer Magazine Luiza, has gained a large following on social media (6.8 million on Instagram). Avatars is also starting to be used for customer service (e.g., NVIDIA Omniverse Avatar), and in digital health services.
The NFT craze in 2021 showcased the potential of avatars being collectibles. Through PFP (profile picture) NFTs, an alternate identity can be tokenized and exchanged among anonymous users. Additionally, for many “blue chip” NFT projects (e.g. Bored Ape Yacht Club), there are also exclusive IRL (in-real-life) events open to NFT holders, further bridging the gap between digital and physical identities.
Across the horizon, we’re also seeing brand new use cases for avatars powered by generative AI (specifically conversational AI), including virtual companions (e.g., Replika) and AI-powered non-playable characters (NPCs) in games (e.g. Inworld AI).
AI and Avatars
Generative AI helps to solve the several current problems for avatars (and by extension, video games in general), including user adoption, interoperability, and monetization
A mantra among gaming VCs that I often hear is that “everyone will be a gamer”. I have a more personal mantra, that everyone will become a denizen of virtual worlds. But how do we get there? I think there’s a good chance that avatars will be the tip of the spear.
There have been many avatar-based social networks over the years, from Second Life and Club Penguin in the early 2000s to IMVU, Zepeto, and Avakin Life in mobile and VRChat more recently. What differentiates these experiences from sandbox gaming platforms such as Roblox, Fortnite, and Minecraft (proto-metaverses) is their primary focus on social interactions (specifically pseudonymous role-playing) in lieu of gaming. These avatar social networks never grew beyond ~10M active users, significant but still niche compared to broader social networks13.
To achieve real critical mass (billions of users), digital identities may have to incorporate more of one's real-life identity14, which also require realistic avatars (the form must fit the function), and that’s where generative AI can help:
Accelerate the avatar creation process
While traditional in-game avatar creation systems relies on the use for the heavy lifting of customization, generative AI takes care of the customization for you in seconds. The algorithms can be trained on large datasets of images and videos of real people, and then be used to generate avatars that closely resemble specific individuals. Ready Player Me is a step in the right direction.
Make avatars hyper-personalized, more than “discrete” avatar creation systems ever can
With the right training data, generative AI can be a massive boost to avatars personalization, adding imaginative, artistic touches while keeping avatars authentic and individualized. Due to the ease of creation, the user can iterate through many creative outcomes rapidly.
Create new use cases to avatars, incentivizing user adoption further
With hyper-personalization and on-going refinement, generative AI has the potential to significantly enhance and transform the way that people create and interact with avatars and virtual identities. Your avatar may “live” even if you’re offline. We are already seeing innovations on personalized avatars in gaming, specifically NPCs. Inworld AI, for example, is an AI-powered tool that allows you to create unique characters with an evolving personality as you interact with them. With generative AI, we can imagine a much more engaging Second Life.
The question is whether social networks will become virtual worlds first, or gaming virtual worlds will grow to have greater mass appeal. Our bet is on the latter.
Interoperability (or Lack Thereof)
There’s been a lot of focus on interoperability of digital assets, mostly from folks who have not spent a lot of times building in Unity or Unreal. The actual process of making digital asset work in different game worlds is often tedious (and may not even make sense outside of its native environment).
Additionally, users don’t necessarily want to look exactly the same in every virtual worlds, and game developers may not want a character running around that doesn’t fit the environment of their meticulously-crafted world. Completely interoperable avatars remove of the primary benefits of virtual worlds - that “you can be another version of you”. Generative AI can help bring some of the benefits of interoperability (ease of use, scarcity) without its imitations.
In-game cosmetics (already a ~$40B market) is one of the lower hanging fruits for generative AI in games. We have already touched upon this in our prior post, so we’ll be brief here.
However, personalized avatars can also become a new source of monetization. For the majority of free-to-play games today, only 5-10% of users ever pay. Lensa has proven that many users are willing to pay for hyper-personalized digital images. Imagine custom avatars created in the theme of Skyrim or Valorant, that is functional with the game’s existing 3D assets and fits with the theme of its existing game world. This can help reduce developers’ reliance on whales to sustain in-game economies, by turning gamers into content creators (at least for their own avatar).
In summary, personalized avatars, and the shift from a simple controller function to an identity function carries with it an increase in monetization potential.
The Future of Avatars
A couple of predictions on the evolution of avatars:
Avatars will start to blur the boundaries between gaming and social media
As social media avatars become more widely adopted and as gaming avatars become more realistic, we see a convergence between gaming and social media avatars:
Personalized avatars will be a major driver of virtual world adoption
Personalized avatars will help drive real-world identity adoption within virtual worlds (“it looks like me, so "I’ll act like me”). This means larger audiences for games, and additional monetization opportunities for social networks.
Generative AI will be an important part of the avatar creation process
Instead of a cumbersome avatar creation system based on discrete variables, we envision the future avatar creation process (and NPC creation process) to leverage generative AI, and potentially look something like this:
Perhaps in 10 years we will not be logging into internet services with our Google single-sign on, but rather our avatar (or our user data repository). The upside for game developers is massive. Even the largest game worlds today are in the 10s or hundreds of millions of MAUs, not billions as we see in the largest social networks.
Opportunities for Startups
There’s a lot of work to be done before the vision we paint is realized. Right now generative models support 2D asset generation, but we (and many others) are awaiting production-ready 3D asset generation.
A couple of ideas we are looking for (not exhaustive):
B2B platforms to help game developers create custom fine-tune models (for avatar generation, and general in-game asset creation)
Smarter NPCs leveraging advances in large language models (LMM)
User data repository for generative AI applications (this may eventually become the “single-sign on” for the digital world)
Next-generation social-first games leveraging generative AI technologies
TIRTA Ventures plans to actively support companies pushing the forefront of AI-enabled avatars and virtual identities. If you’re building in this space, do not hesitate to reach out.
About TIRTA Ventures
TIRTA Ventures (“TIRTA”) is newly launched VC firm based in New York City and focused on the interactive entertainment industry. TIRTA was founded by Ben Feder (former CEO of Take-Two Interactive, former President of Int’l Partnerships at Tencent Games).
TIRTA invests across content (studios), infrastructure/technology, and platforms that support the broader gaming ecosystem. We seek to bring our decades of operational experience in the gaming industry to our portfolio companies.
For more information, please visit our website (under construction).
Lensa’s photo generation feature was built on top of Stable Diffusion, a foundational model. Applications like Lensa of often build their own “fine-tune model” to customize the underlying foundational model’s various parameters to fit a particularly use case.
I tend to use digital worlds and virtual worlds synonymously, although one can argue that virtual worlds is a subsect of digital worlds - more akin to the popular view of “metaverse”.
In experiments, players who were given more attractive avatars acted more confidently and were more outgoing than those with unattractive avatars.
The non-gaming digital avatar market is estimated to be $10B in 2020. Source: Emergen Research.
The need for the “identity” function is greatest among RPGs and sandbox games.
Readers born after the early 90s will never appreciate the frustration of dial up internet, and the joy when it finally connected.
Aside from the username.
Internet users in East Asia show a preference for non-realistic profile pictures or avatars, even today.
Few people will recall Trutoon, a 2008 startup that attempted to create personalized avatars through crude photo-shopping (pose with Obama was their claim to fame).
Gorillaz is primarily led by Damon Albarn, the frontman of the English rock band Blur. Damon’s identity in Gorillaz takes the form of 2-D, allowing Damon to experiment with new musical styles, essentially establishing an alternate identity through his avatar.
Donna Li, The Rise of VTubers: An Overview of the Surging Popularity of “Virtual YouTubers”, https://thesciencesurvey.com/arts-entertainment/2022/07/24/the-rise-of-vtubers-an-overview-of-the-surging-popularity-of-virtual-youtubers/
Andrew Amos, VShojo announces $11 million funding round as VTuber agency expands, https://www.dexerto.com/entertainment/vshojo-11-million-funding-round-vtuber-group-expansion-1784541/
Although MMORPGs were some of the first virtual worlds, the “RPG” element inherently made these experiences niche - not everyone wants to role-play. Role-players certainly found significant value in digital avatar worlds, but there’s a good chance that use-case remain niche in the future.
There’s a spectrum of identity realism. Given the appeal of pseudonymous gameplay today, perhaps not this system can be opt-in, or restricted to certain aspects of the identity, e.g. appearance, voice, etc. if not name itself.