Avatar

The LP1 (2018)

The first version of the Low-Poly avatar project was started in 2018 as a potential replacement for the SHL3 avatars. The SHL3 avatars were your typical blocky Roblox avatars, with the main difference being a proportion adjustment that made them a bit more humanoid scaled, even if they were still composed of rectangular blocks. Ryferion1 made a basic segmented character mesh to allow us to do some basic overlay testing for the Roblox avatar.

We ended up with a functional (at the time at least) prototype that could walk casually across a baseplate, but ended up not feeling confident enough in it to warrant remaking all the assets in SHL3 to fit it. It was mostly unused in future iterations, though the LP2 did grab a library from it that allowed for easy manipulation of the bones of the Roblox avatar.

The LP2 (2019)

For the summer project Mortal Metal, the LP2 was unleashed. It was overlayed upon an invisible general Roblox avatar, saving a lot of time. To avoid having to make a bunch of hats and figure out uncanny faces, we kept the Roblox head, creating a hybrid style between the two. People liked the aesthetic, but weren't a fan of the fairly limited customization compared to the SHL games. Unlike previous games the outfits were almost entirely prefab, acting as things to buy rather than ways to express yourself.

The game was a third person shooter, so first person was never fully integrated.

Limbs were segmented and often separated during animations due to no mesh deformation technology available in 2019

While there was general size scaling, body proportions were baked into prefab outfits.

The NPC enemy Necroborg design didn't match any of the scaling standards due to being created first, with the skull most noticeably being more humanoid.

The LP2 had a few limitations that caused its inevitable decline. The combination of the lower detail Roblox body style with the higher detail outfits was a bit uncanny. On top of that the gluing of it to the Roblox avatar made it difficult to work with, requiring changes to work within both the context of the LP2 but also the general Roblox avatar. Another limitation was that the existing customization system made it difficult to use in SHL sequels. The final nail in the coffin was the debut of mesh-skinning in Roblox, which made the segmented joints of the LP2 simply uncompetitive in the avatar customization field.

The LP3 (2020)

The LP3 is the result of 5 months of iteration and testing. It ditched the Roblox avatar system entirely. It created multiple new frameworks and libraries for future versions, and while not universally liked by the community, made a considerable splash within the general Roblox Developer community at the time for improving upon RThro - the official Roblox brand of realistic avatar.

More Realistic

Mix & Match Style Customization

Hair Physics

Face Scaling

While the LP3 was of course not without downsides, such as its long development period due to multiple code refactorings, the limited number of compatible assets, and of course the instability that comes with constantly refactoring. At no point was there a definitive LP3, with nice features such as scaling never coexisting with polished movement, and nice transformations coexisting with controlled customization. Hair started off strong and got less and less stable until it was replaced last minute. None of these issues were unfixable - in fact given time most would have been repaired and polished - however the death blow came suddenly.

The introduction of the cage mesh deformer, and the recognition that the LP3 would not be able to compete with a basic Roblox avatar within the next year.

The LP4 (2021)

Taking what worked with LP3 and bringing it to the next level using Roblox's new Cage Deformer technology. With this technology you can apply layers of clothing on any body type, even layering over other pieces of clothing! It's incredible, and if SHL4 is to compete against other games using it, it also has to incorporate it.

Through using this new technology, it frees up a commitment to a single avatar style. Instead we can lay the same outfit onto every possible body style, something the player could set on their client. This means everyone you see will have the same avatar-body style as you, leading to the feeling of an actual consistent style in the game. This allows for everyone to see the world as they like without having to sacrifice their preferred character look.

Unlike the LP3, which took almost 3 months before it was playable, the LP4 is limited not by coding but by a combination of asset production and the live release of the new cage mesh technology. This means that the actual amount of time spent programming it will likely be in days, rather than months. Due to the limited number of assets created for the LP3, it is expected we'll be able to recreate faithfully most characters within LP4 automatically. It will likely save time long term, as we intend to use the R15 bone format to allow for recycling of R15 animations, as well as the overall increased modularity of assets.

Here are some of the potential body styles you will be able to choose from. Each has its own abbreviation for clear referencing purposes, though in the final game they will likely be chosen from a visual list, rather than a list of names. You'll be able to determine what sections of faces show up (eyes, noses, mouths, ears, etc), as all of the facial features will be additive rather than baked into any particular body.

LPI: Indie

This style takes what we felt worked with the LP3 and improves upon it, as well as takes some extra steps to further abstract away from actual people. We do feel LP3 proved the potential for more realistic avatar, but the amount of time it takes to polish such a look proved to be very time consuming - getting the face and hair right ended up taking an estimated 2 months.

The final style is inspired by the looks of various indie development studios over the years. The low-poly indie look is one that many will recognize, but hasn't really made its way onto Roblox yet.

Aesthetic

The design will be similar to the flatter, more angular Polygon Avatar Style.

Face

The faces will be more cartoony, with parallels to the Akishaqs Avatar Style. Unlike with the LP3, face details will be applied over the base face mesh. It is uncertain if noses will be included at this time.

Hands

We will be experimenting with actual low-poly fingers on the avatar, similar to those shown in hands made by an artist named Spug. These hands will possibly rigged for basic hand clenching motions, otherwise they'll likely default to the semi-closed position frequenting Roblox.

LPR: Roblox

This version won't be a direct copy of the Roblox avatar style, especially since there isn't one definitive style anymore. It will however be less jarring to players who are familiar with the Roblox aesthetic, and aren't quite ready to lose it.

Aesthetic

The look will take large inspiration from the Roblox 3.0 body type, with the notable differences of changing the hands, connecting the shoulders to the arms smoothly, and making the feet a bit blockier. The avatar will be a bit on the leaner side compared to Roblox 1.0 unscaled, however we feel this is necessary to have it work with assets tailored for LPI.

Face

Flat face meshes will be used to mimic the decal based faces on Roblox, with some standardizations around the dot based eyes, as well as simpler non-abstract mouths. There won't be any noses.

Head

We won't be fully embracing the generic round head style for the body. Expect some deviations from the general shape, similar to the magnitude changed as with the "Man Head" used in a Roblox package. It's obviously still Roblox, but just looks a bit more like a human skull instead of a Lego minifigure.