The gap between 2D passthrough porn and volumetric VR porn is wider than most people realise — not just in how they look, but in how they’re made, how they render, and what your brain does with them. If you’ve used one and are curious about the other, or you’re deciding where to send your subscription money, this is the full technical breakdown.
If you’re new to passthrough VR porn entirely, start with the What Is Passthrough VR Porn explainer first — this article builds on that foundation.
Disclosure: By clicking on product or external site links on this page, we may earn a small commission—at no extra cost to you. Read the full disclosure.
2D passthrough uses a flat video of a performer, shot on green screen, composited into your room’s live camera feed. She appears in your space — at full human scale, in your environment — but as a flat image. A life-size cutout, not a physical presence.
Volumetric VR porn captures the performer as actual 3D spatial data. She exists in your room as a 3D object. You can move around her. Step left and you see her from a new angle. Step closer and perspective shifts like it would with a real person standing there. This is not a different version of the same technology — it’s a fundamentally different approach.
The production pipeline is well-established: green screen shoot → chroma key removal in post → playback via DeoVR, PLAY’A VR, or HereSphere, composited over your headset’s real-time camera feed. The passthrough explainer covers the full production and playback chain in detail.
What matters for this comparison: the output is a video file. Flat, 2D, played back at a fixed camera angle. The intimacy comes from placement — she’s in your room — but your spatial relationship to her is static.

Volumetric VR porn is not a video. It’s a spatial reconstruction of a real person — a mathematical model that stores the geometry and appearance of a performer from every possible angle simultaneously.
When you load volumetric content in your headset, you’re not playing back a recording. You’re rendering a 3D model built from a real person, in real time, from your exact viewpoint in space. Move and the render updates. The perspective is genuinely yours, not predetermined by a camera operator.
The technology currently powering all commercially available volumetric porn is Gaussian Splatting.
Gaussian Splatting is the piece of this that’s worth understanding properly — because it explains everything about the experience, including its current limitations.
The core idea: instead of representing a 3D scene as a mesh (like a video game character) or as pixels in a video, Gaussian Splatting represents it as millions of tiny translucent ellipsoids — called “splats” — each with a position in 3D space, a colour, an opacity, and a shape. Rendered together, these overlapping ellipsoids produce a photorealistic-looking result from any viewpoint.
Here’s the capture-to-render pipeline:
To build a Gaussian Splat of a performer, studios surround them with a rig of 50 to 200+ synchronised cameras firing at exactly the same instant from every angle. This is not a green screen stage — it’s a dedicated capture stage where nothing moves except the performer.
The image array from every camera feeds into photogrammetry software that estimates the 3D position of thousands of points on the performer’s surface. The Gaussian Splatting algorithm then places ellipsoids at each point, iteratively optimising their colour, opacity and shape until the rendered result matches every camera view simultaneously.
This optimisation process takes hours — sometimes days — of compute time per second of captured footage. That’s why you get loops, not scenes.
A standard 8K VR scene might be 20–30GB for 30 minutes. A 10-second Gaussian Splat loop at Braindance runs around 15GB per model — because instead of compressed video frames, you’re storing millions of 3D data points with spatial position, colour, opacity and orientation for each one. That’s not a compression problem; that’s what spatial data costs. Splat compression is an active research area and will improve, but it’s a genuine constraint right now.
In standard VR porn — including 2D passthrough — the camera position is locked. The video was shot from one angle, and that’s the angle you watch from. Head tracking lets you look around, but your position relative to the performer doesn’t change based on where you physically stand.
With Gaussian Splatting, the renderer calculates your exact headset position in real time and renders the splats from that viewpoint. Step two feet to the left and the render updates to show her from two feet to the left. Move closer and perspective shifts. The occlusion — the way she blocks what’s behind her — works like real physics because it is real geometry.
Before Gaussian Splatting became the standard, NeRF (Neural Radiance Field) was the leading volumetric approach. Some platforms still reference it, and it’s worth understanding the difference.
NeRFs work differently: instead of discrete splats, a neural network learns to predict what a scene looks like from any viewpoint. Given a position and direction, the network returns a colour and density value. The entire scene is stored implicitly — inside the weights of the neural network, not as geometric data.
| NeRF | Gaussian Splatting | |
|---|---|---|
| Render speed | Slow (seconds per frame) | Fast — real-time capable |
| File size | Smaller (network weights) | Large (15GB+ per model) |
| Sharp surface detail | Good at fuzzy/translucent | Better at crisp surfaces |
| Standalone headset use | Impractical | Yes — runs on Quest 3 |
| Current use in porn | Rare | Dominant |
| Who uses it | Research / niche platforms | Braindance, Real Girls Now |
Gaussian Splatting has largely replaced NeRF for real-time playback because it renders fast enough to run on standalone headsets. NeRF still has theoretical advantages for certain fine details and static scenes, but the render speed problem makes it impractical for consumer porn apps.
| 2D Passthrough | 3D Volumetric (Gaussian Splatting) | |
|---|---|---|
| Performer depth | Flat (2D) | True 3D object |
| Walk around her? | No | Yes — full parallax |
| Scene length | 30–45 min full scenes | 10–20 sec loops |
| Library size | Hundreds of scenes | Small (handful of models) |
| File size | 20–30GB per scene | ~15GB per model |
| Production method | Green screen + chroma key | Multi-camera capture stage |
| Streaming app | DeoVR / PLAY’A VR / HereSphere | Platform’s own app |
| Headsets | Any standalone | Quest 3/3S, Pico 4 Ultra, PCVR |
Stats don’t capture how different these feel to use. With 2D passthrough, immersion comes from placement. She’s in your room. Your brain supplies the rest. The intimacy is high, the presence is compelling, but your spatial relationship to her is fixed — you’re watching through a window, not sharing a space.
With volumetric, your positions are genuinely shared. She occupies your room as a 3D object. You can crouch and look up. Walk around and see her back. Step close and the perspective shift is physically correct. Braindance specifically designs for this — performers are positioned so the user can simulate intercourse from a natural angle, and because the geometry is true 3D, occlusion works like real physics. Her body actually blocks what’s behind her based on where you’re standing.
The trade-off is content length. A full-length passthrough scene runs 30–45 minutes. A Gaussian Splat loop runs 10–20 seconds. These are not comparable watching experiences — they serve different purposes. To see the more interesting VR porn kind, check our Best Passthrough VR Porn Sites Ranking.
This content is blocked because YouTube cookies have not been accepted.
This technology is early. The limitations are real and worth understanding before you commit money to it:
These are engineering constraints, not design choices. All of them are actively being worked on. The limitations today are not the limitations of the technology — they’re the limitations of current tooling and compute budgets.
For 2D passthrough, the requirement is any standalone headset with streaming app support — Quest 3, Quest 3S, Pico 4 Ultra, or Samsung Galaxy XR. Full headset breakdown in the Best VR Headsets for Porn guide.
For volumetric specifically:
The render load for Gaussian Splatting is meaningfully higher than video playback. Quest 3’s Snapdragon XR2 Gen 2 handles it adequately. Quest 2 can struggle with complex models. If you’re primarily interested in volumetric content, the Quest 3 is the practical minimum for a comfortable experience.
Apple Vision Pro’s display quality would be exceptional for volumetric content — and unlike 2D passthrough, some volumetric apps don’t rely on the passthrough API and may not face the same developer restrictions. Neither Braindance nor Real Girls Now has confirmed Vision Pro support at the time of writing, but it’s less of a closed door than standard passthrough apps face.
The current limitations are mostly compute and tooling constraints. What’s in active development:
The trajectory is clear. The question is timeline.
Neither is universally better — they serve different experiences.
Choose 2D passthrough if:
→ The Best Passthrough VR Porn Sites guide covers every platform worth subscribing to, from SexLikeReal’s AI passthrough to ARPorn’s studio productions.
Choose volumetric if:
→ Braindance is the strongest option for photorealistic volumetric. Real Girls Now is worth trying for longer striptease and dancing sessions. Both are reviewed in the Passthrough Sites guide.
Share this article
Author:
Founder & Lead Reviewer at VRPornJudge
Henry Miller is the founder of VRPornJudge and has been independently reviewing VR porn platforms through hands-on testing since 2019. His evaluations are based on extended real-world use, including full subscription access, long-term content sampling, and performance testing across multiple VR headsets. Henry places particular emphasis on playback stability, content quality over time, and interactive toy compatibility.
Testing equipment: Meta Quest 3, Samsung Gear VR, The Handy (teledildonics device).