• We at present assist two completely different tech stacks for avatars: A legacy tech stack (R6) that helps older avatars and experiences; and a more moderen tech stack (R15) that helps all avatar types and capabilities.
• To make sure that any avatar model will work in any expertise and everybody can entry the newest options, we’re working to unify these right into a single tech stack. 
• This presents technical challenges, so we’re working intently with our developer group to launch instruments that can ease the migration onto the brand new unified tech stack.

Avatars are more and more changing into part of our identification. At Roblox, we wish every of our greater than 65 million day by day customers to have an avatar that they really feel actually represents them—not solely how they give the impression of being, but additionally how they categorical themselves to others in actual time. This turns into much more necessary as we launch immersive communication instruments like Join, which is a brand new approach for anybody 13 and older to name mates on Roblox as their avatar. For folks to really feel actually linked as their avatars, they want to have the ability to react and present emotion within the second. We want avatars able to extra advanced facial expressions, lip syncing to voice, and nonverbal cues, comparable to shrugging or nodding. 

To make sure that everybody can see themselves mirrored in these immersive worlds, we’ll want a higher number of components that folks can combine and match to make avatars that symbolize them. Which means extra physique and head varieties to select from, in addition to extra clothes, make-up, and accent varieties, and extra hair and pores and skin colours, textures, and types. To quickly increase the alternatives for these things, we’re working to make it a lot simpler to create new avatars and empower extra folks to convey their concepts to life. We’ve come a great distance since our first blocky yellow avatar, and we aren’t completed but. 

As avatars evolve and enhance, we additionally need to make sure that the newest developments, together with layered clothes, facial animation, chat with voice, animation packs, and emotes, can be found for each avatar, in each expertise. At present, solely avatars constructed on our most trendy tech stack—known as R15—have entry to the newest mobility and expression capabilities. That’s as a result of we at present assist two distinct avatar tech stacks. The R6 tech stack was designed for the basic blocky-style avatars, which have solely six physique elements, and the experiences constructed for these avatars. The R15 tech stack was designed to assist avatars with as much as 15 physique elements, so it helps all avatar types—blocky, humanoid, and fantasy—and experiences constructed for all avatars. Supporting twin tech stacks has created limitations and frustrations for builders and creators. 

We at present assist greater than 15 years of experiences, a lot of which had been designed for R6 expertise and will not be working as seamlessly with the most recent, most expressive avatars as we’d like. For instance, if somebody with an avatar constructed on R15 enters an expertise constructed on R6, their avatar might look and transfer in another way than common—their avatar would not have the ability to make facial expressions. If they’d layered clothes, comparable to a jacket over a shirt, their avatar would revert to less complicated clothes. As well as, some experiences, like impediment programs, are constructed round particular avatar sizes. We all know this isn’t supreme for individuals who use or create for Roblox. 

We wish everybody on Roblox to have entry to our most superior avatar expertise to allow them to totally embody their digital identities and create wonderful experiences and visuals. We additionally need to be backward appropriate with current avatars and experiences. Given all of this, we’re being very considerate about how we strategy this unified tech stack, to keep away from creating additional disparities and to create a path ahead that minimizes the quantity of guide work required. We are going to present the builders constructing these worlds with the instruments and assist to maintain their experiences vibrant and interesting whereas sustaining the texture they need for his or her expertise. 

Shifting to a unified tech stack

Our avatars—blocky, humanoid, or utterly fantastical—ought to simply work in any expertise, with any accent. We need to take away any friction creators and customers have felt thus far. We additionally need creators to retain management over the appear and feel of their experiences, whether or not they assist R15 tech, or R6. To assist all of those new options and capabilities—now and as we proceed to innovate—we’re unifying the technical structure that helps all avatars.

We’ve heard from our developer group that they need to maintain the appear and feel of the basic blocky avatar model, however in addition they want us to implement constant avatar sizes and proportions. We additionally heard that they need instruments to make it simple to load avatars constructed on R15 tech into R6 experiences now—and the power to automate the method of changing R6 experiences to R15 requirements. Our longer-term objective is to construct a layer that can allow R6 experiences to work with the R15 stack, whereas minimizing any specialised code we would wish to keep up.

Earlier this 12 months, we shared the R6 to R15 adapter. The adapter works as an emulation layer, permitting R6 scripts to run on R15 our bodies, with out requiring any motion on the a part of the avatar’s creator. When an R15 avatar joins an R6 expertise, the adapter permits it to maneuver in the identical approach as an R6 avatar. This enables builders to instantly check out R15 avatars with only one click on and see how nicely they work earlier than making any updates to their experiences. With this new adapter, R15 avatars retain options like layered clothes and facial expressions, however can nonetheless be part of an R6 expertise and transfer because the developer initially supposed.

Our subsequent step can be a set of conversion instruments to permit builders to simply migrate their R6 experiences to the R15 tech stack. These instruments will assist builders convert an expertise’s script, character, and animations and assist them check the conversion as they go. The conversion instruments will use the R6 to R15 adapter so builders can publish their experiences in the course of conversion with out breaking. Lastly, we plan to provide builders the power to regulate avatar scale to any desired setting, together with mirroring the basic Rthro avatar model. This provides builders consistency for impediment programs and unlocks the potential for constructing new varieties of Roblox experiences. 

Past the unified avatar tech stack

Migrating to a unified tech stack is a mandatory step for us to assist builders and customers as we enhance avatar expertise and introduce new options and instruments. But it surely’s just the start. Unifying all avatars on one tech stack will make it simpler for builders to benefit from new real-time communication instruments, comparable to Join. For these calls to really feel like a pure dialog, we’ll all want entry to newer avatar capabilities like facial expressions, emotes, and voice syncing. We additionally need to allow a much wider number of avatars so we not too long ago opened the doorways to avatar creation by any of our UGC members. We additionally introduced that we’re engaged on a generative AI instrument to allow anybody on Roblox to simply create an avatar from a picture and a textual content immediate. 

Our objective is all the time to be a platform that connects folks with security and civility in thoughts, so we’re considerate about how we’ll average the creations and interactions with these new avatars. As instruments like generative AI democratize and speed up creation, our moderation efforts have to maintain tempo, leveraging a mix of AI and human moderators. A few of the challenges that we’re at present addressing are straight associated to the combinatorial nature of avatar creation and the huge variety of social interactions on the platform. We’ll share extra particulars about our moderation instruments as we launch them. 

Finally, we intend to allow anybody to create and customise avatars from scratch—even from inside an expertise. This can unlock limitless methods for folks to precise their individuality. From a technical and creator standpoint, in addition they current various attention-grabbing technical challenges to unravel:

• How does a creator design objects for an enormous array of avatars with no restrictions to physique symmetry, variety of limbs, or facial options, whereas additionally supporting options like layered clothes or the power to animate the avatar’s facial options?
• How can we allow extra folks to create avatars with out having to make use of skilled 3D graphics software program?
• How can somebody’s personalised avatar match seamlessly into any expertise they discover on Roblox? 
• With the fast proliferation of UGC avatars and highly effective generative AI strategies, how can our groups optimize our grid and cloud for max stability, in addition to low latency, and effectivity?

We’re working to unravel these challenges with new instruments for creators, new infrastructure to make the platform much more dependable, and persevering with to speak transparently with our creator group. By getting everybody onto one unified tech stack, and releasing instruments to make all of this simpler, our creators will have the ability to do what they do greatest: Blow our minds by creating issues we by no means may have imagined.