P_01
Resolution
More data, more better.
Samples carry significantly more information than pixels — the raw material for upscaling, denoising, AI processing, and everything downstream.
Ver_0.9 // Sub-pixel Sample Engine
EVOTIS® — a GoGhost product
The end
of pixels.
For thirty years, every rendering pipeline has thrown away the most valuable data in your image. We built the software that stops deleting it.
09
US patents
∞
Samples per pixel
0
Edge artifacts
1
File. All views.
01 // heresy
Pixels
suck.
For the past thirty-plus years, every image processing pipeline in visual effects, broadcast, and rendering has ended in the same place: a grid of pixels.
That made sense when monitors, TVs, and phones were the whole universe. It doesn't anymore. Displays evolved. Cameras evolved. Renderers evolved. The pipeline in between them didn't.
It's hard to imagine a world where you can separate pixels from the display. Allow us to open your eyes to the world of non-uniform samples.
Then
Fixed pixel grid, thrown-away data
Now
Non-uniform sub-pixel sample cloud
Result
Resolution-independent, render-accurate
Cost
Less rendering, smaller teams, more art
02 // the waste
Samples?
A renderer, a sensor, a lidar — none of them produce pixels. They cast rays from the viewer into a scene and record the answer as non-uniform sub-pixel samples — irregular measurements carrying color, depth, motion, and every AOV in between.
At the very end of the pipeline those samples are averaged into a grid of pixels — and then thrown away. Everything precalculated: gone. All that's left is one blended number per cell.
This was a compromise for 1980s memory and 1990s bandwidth. Neither is a problem in 2026. Raster is a dead format for high-fidelity work.
ARays cast from viewer
BSub-pixel samples land
CAveraged → samples deleted
fig_02 // legacy raster pipeline→ evotis keeps step B alive, forever
03 // the breakthrough
Time for a change.
Through years of engineering, GoGhost found a way to efficiently store and process non-uniform sample data — enabling workflows for artists and technical directors that were impossible in a traditional pixel pipeline.
Nine
US patents covering sample compositing, encryption, and streaming.
Live
Streaming from Redshift today. Nuke plug-in available.
Silent
The renderer never has to know what it's rendering.
04 // see it move
Live from the pipeline.
Four captures from EVOTIS running inside Nuke and streaming from Redshift. Hover to play.
DEMO_01
Nuke plug-in
Interactively create arbitrary pixel resolution (upscale), generate render-accurate mattes, and visualize the underlying samples themselves.
DEMO_02
Redshift streaming — interactivity
Two EVO files loaded like DEEP. Move elements interactively with render-accurate depth filtering, edges, and preserved render passes.
DEMO_03
Optically-correct lightwrap
HDR background at 1 sample per pixel. Foreground composited entirely in sample space for physically correct lightwrap.
DEMO_04
Volumetric composite
iPhone still with 1 sample-per-pixel depth. Redshift-streamed geometry volumetrically composited into the plate in real time.
05 // what it unlocks
Three doors,
all open at once.
P_02
Security
End-to-end, renderer-blind.
Encrypted sample streams mean the render farm, subcontractor, or cloud vendor never has access to the finished image. Not even the renderer decodes it.
P_03
Workflow
Ship in fewer round trips.
Creative decisions that once forced a re-render happen at sample level. Dramatically fewer trips back up the pipeline. Smaller teams, more control.
06 // benefits for M&E
What breaks
in your favor.
Every one of these is a compositing or pipeline problem your team spends real hours on. All six collapse when the image is stored as samples instead of pixels.
benefit_01
Arbitrary matte generation
No setup in 3D. Render-quality mattes post-render. No edge artifacts, even through motion blur and depth of field.
01benefit_02
Grading at sample level
Kill fireflies from CG renders. Recolor individual objects without breaking edges between them.
02benefit_03
Resolution independence
Resolution isn't decided before render. Draw any pixel grid you want over the samples. Append more samples per-object as needed.
03benefit_04
Deep compositing killer
DEEP is pixel-space with depth. EVOTIS is non-uniform, real-time, and unlimited X/Y/Z in the same space.
04benefit_05
Non-destructive compositing
Process samples directly. Filter to pixels only at display time. New real-time workflows that make rasterized compositing obsolete.
05benefit_06
Stereoscopic & VR
Fill 'shadow' areas from perspective shifts. Avoid rendering what's already filled. Composite every view as a single file.
0607 // in plain english
For the people
reading the P&L.
Content is expensive because processing time and human hours are expensive. Storing the sample data instead of throwing it out is how you buy those hours back.
- cost_of_labor
Storage is cheap. Talent isn't.
Cloud storage is one of the cheapest line items now. Processing time and skilled labor are the ones eating your budget every year.
- linear_flow
One flow of data, no re-renders.
Artists and TDs get downstream control. Fewer trips back up the pipeline for another render pass. Time back, money back.
- smaller_teams
Smaller teams, more shipped.
Proxy workflows stop being throwaway. Everything you build moves toward the final product instead of being tossed.
- spot_compute
Cloud workflows, cheaper compute.
Sample-native pipelines open up spot compute — the cheapest way to render at scale. The math finally works.
08 // encrypted imagery
The renderer
doesn't need
to know.
Sample rays are scrambled at cast time using public/private key encryption. The renderer, farm, and every subcontractor in the chain process the samples without ever seeing the image.
Scrambling a pixel image doesn't work — someone had access to the ordered samples to make it. That's a leak vector. Doing it at the ray-cast level closes it.
Absolute security from start to finish. Even the renderer can't decode it.
U.S. Patent No. 10,127,392 B1
Secure rendering system that generates ray-tracing samples with obfuscated position data.
Read the patent→sample_stream // scrambledrt_cast
POS_ENC
19SYouJY4RtGqyJU3UN…POS_ENC
a7c9F0zXqB1p4RtGqyJ…POS_ENC
Kf4nMzZ0oLC8AR9U7Fr…POS_DEC
(1284, 720) → RGBAkeys held client-side. never leave the studio.
pipeline // obfuscation
Positions are scrambled before the ray is cast.
Per the patent: each sample carries color and auxiliary data, but its position on the image plane is transformed by a key-driven obfuscation function. Rendering proceeds normally on scrambled coordinates. Only a holder of the inverse key can place samples back on-canvas.
- 01Studio
Generates key pair. Public key ships with the render job; private key never leaves.
- 02Ray cast
Sample position (x,y) is transformed by the obfuscation function before the ray is fired.
- 03Farm
Renders on scrambled coordinates. Color, depth, AOVs — all correct. Layout — unknown.
- 04Return
Encrypted sample file is streamed back. Interceptors see noise, not imagery.
- 05Descramble
Private key reorders samples on-canvas. First and only cleartext frame lives inside the studio.
What we defend against
- Rogue operators inside render farms
- Compromised cloud storage or CDN cache
- Subcontracted VFX vendors and sub-vendors
- In-transit interception of dailies & masters
- Screenshots of monitored review sessions
- Post-render frame leaks pre-release
What the patent enables
- Key-driven permutation of sample x,y positions
- Symmetric or public/private key modes
- Multiple keys for tiered access (previews, finals)
- Selective descrambling — region, resolution, layer
- Watermarking via key fingerprint per recipient
- Zero cleartext frame ever exists on the farm
00
Frames rendered in the clear
1:1
Sample → key mapping
∞
Recipient-specific keys
09 // resolution independence
One cloud.
Any grid.
Resolution isn't baked into the render. The samples exist once — the grid is a filter you draw over them at display time. Drag the slider to redefine the pixel grid over the same 98304 samples. Coarser = more samples per pixel = quieter image. Finer = higher res, fewer samples per cell. Not enough? Append more. No re-render.
samples + grid overlay
rasterized image · 16×16
resolution16 × 16 · 256 pixels
481632486496128192256
98304
samples
16×16
grid
384.00
samples / pixel
At 16×16, each pixel is reconstructed from roughly 384.00 samples on average. Push resolution higher and the image softens — the samples spread thinner across more cells. Append more samples and it sharpens again. No re-render.
10 // defensibility
Nine patents.
One idea.
The full portfolio covering sample-native compositing, encrypted rendering, streaming, and multi-view generation. Filed and granted over years of engineering.
- P001US 9,990,761Method of image compositing directly from ray tracing samples
- P002US 10,127,392Secure rendering system that generates ray tracing samples with obfuscated position data
- P003US 10,410,001Video security and ray tracing samples compression system
- P004US 10,372,920Non-rasterized image streaming system that uses ray tracing samples
- P005US 10,410,433Method of modifying ray tracing samples after rendering and before rasterizing
- P006PCTMethod of modifying ray tracing samples (continuation)
- P007PendingMulti-viewpoint image generation system
- P008PendingSecure rendering system that generates ray tracing samples with obfuscated position data (continuation)
- P009PendingRendering and filtering options
11 // request access
Stop shipping
pixels.
We work directly with studios, broadcasters, and image-security teams. Tell us about your pipeline and we'll set up a live demo.
$ current integrations:
→ Nuke plug-in (available)
→ Redshift live streaming (available)
→ Custom pipeline (contact)