Gracia’s Hands-On Experience with Volumetric Scenes: My Impressions

At AWE Europe, one of the standout moments for me didn’t even take place on the official show floor. It happened in the event corridors during a spontaneous demo of Gracia’s moving volumetric scenes. Let me share why this experience truly captivated me.

Gracia is a startup with an ambitious goal: to become the YouTube of volumetric content, utilizing Gaussian Splats for rendering. Their vision is to allow creators to upload 3D scenes—perhaps using just their phones—onto Gracia’s servers. From there, viewers can dive into these scenes solo or with friends, experiencing them as the creator intended.

Right now, Gracia is just starting out. They’ve released an app for both Quest and Steam VR, where you can explore a handful of scenes scanned by their team. The Quest version is pretty basic; you’re limited to a set few scenes. However, the Steam VR version opens up opportunities to explore additional scenes scanned by users. Because Gaussian Splats are resource-heavy, the Quest version requires simplified visuals, which affects quality. Still, it’s remarkable that it functions on Quest at all. Gracia’s team has worked hard on optimizing the experience for this platform.

A few weeks back, I checked out Gracia on Quest and found the platform promising despite some hiccups, like visible scan artifacts and a clunky UI. Unfortunately, I couldn’t compare it to the PC version at the time due to a lack of suitable hardware.

Then, at AWE, Tipatat Chennavasin, an investor, came through. He surprised me with a demo right at the venue, using Gracia on his high-end laptop to showcase the superior performance on PC, especially highlighting the brand-new feature: moving volumetric scenes.

Gracia’s team has now managed to digitize several short animated scenes—each lasting between 10 and 30 seconds and usually involving people. With the SteamVR app, you can enjoy these as brief volumetric displays unfolding before you, still powered by Gaussian Splats. This is a pioneering step because similar platforms have typically only managed to display static scenarios. According to Upload VR, however, there’s a hefty price to pay for this innovation; these scenes are taxing to scan and render, hence they’re still in the experimental stage.

Gracia admitted that capturing these clips required pricey 3D scanning setups, with each individual frame taking six minutes to train. In parallel, this process could still take weeks per minute of footage, given the high AWS compute demands. Plus, these clips demand more than 300 megabytes per second to download, theoretically requiring a 2.4-gigabit internet connection, which isn’t common yet. Fortunately, Gracia anticipates substantial improvements in compression technology over the next year, which should make such content more accessible via modern fast internet or through cloud VR streaming solutions.

During the demo, Tipatat set up Gracia and handed me a Quest headset linked to his PC with an eye-catching orange headband and cable. The experience began in the familiar UI of Gracia, which I quickly bypassed. Before long, there stood a blonde girl, dancing right before me, leaving me truly awe-struck.

This revelation in visual quality was striking. The Quest version, while decent, exhibited some over-compression, leading to slight artifacts that gave the appearance of fur on the character’s skin. But on PC, these issues vanished, leaving me with a much more authentic portrayal. Although our brains can detect even minor imperfections, and not every element was flawless (some parts seemed brush-like), the realism was astounding—superior to a Metahuman.

What truly blew me away was her fluid movement. Watching her dance, and being able to walk around her and observe the scene from every angle, felt astonishingly lifelike. At one point, I crouched for a different perspective, only to incite laughs from Tipatat accusing me of peeking under her skirt. Despite my protests, everyone shared a good laugh.

This amusing moment highlighted the potential of such technology in the adult industry, ripe with opportunity to leverage this level of interactive realism. Adult entertainment often pioneers advancements, and this could stimulate further innovation in Gaussian Splat technology.

Adding to the fun, I could alter the size of the dancing figure using the controllers, shrinking her down to fit in my hand, where the smaller scale masked the brushstroke artifacts, making her seem even more lifelike.

The video paused to allow close inspection, revealing the detailed semi-transparency of her attire. Following this, I explored two other scenes: a heartwarming portrayal of parents dancing with their baby and a chef leading a cooking tutorial. While the first evoked genuine happiness, the latter, though well-executed, was hindered by more artifacts and felt somewhat sterile.

In summation, the moving volumetric scenes from Gracia were far beyond my expectations. They hint at a future where people might capture life’s moments in 3D scenes to revisit whenever they wish. Though widespread accessibility isn’t imminent—tools remain expensive and heavy—what I experienced felt close to consumer-ready. The leap to simpler, more affordable technology for capture and rendering will propel us toward a new era of 3D memories.

For now, if you have a powerful PCVR setup, I highly recommend experiencing Gracia’s moving scenes—they’re genuinely impressive.