Upscale
Alpha Real-ESRGAN upscaling: 8192×4096 native, up to 16384×8192
Alpha: Real-ESRGAN and Projection-Aware Upscale are available for testing and feedback. Their workflow can change before production release.
Upscale uses Real-ESRGAN, an AI upscaling model, to enhance your panoramas while accounting for the equirectangular projection. On an open render, click Resize and choose a target in its Upscale section. Transform 4K images into rich 8K or 16K outputs, up to the 16384×8192 application cap.
Resolution Scaling
4096×2048
8 Megapixels
8192×4096
33 Megapixels
16384×8192
134 Megapixels
Key Features
Neural Upscaling
Real-ESRGAN uses deep learning to intelligently upscale images, adding convincing detail and texture rather than simple interpolation blur.
Detail Enhancement
Enhance textures, sharpen edges, and recover fine details lost during compression or lower-resolution rendering.
How Projection-Aware Upscale Works
One sphere, two projection regions
A flat-image upscaler treats every row alike even though equirectangular pixels stretch toward the poles. Projection-Aware Upscale keeps the mid-latitudes horizontally wrap-aware, processes north and south through stereographic polar caps, then blends the regions into one equirectangular result.
-
1
Wrap-aware mid-latitudes
The panorama's left and right edges remain one continuous join
-
2
North polar cap
The zenith is processed through a stereographic cap
-
3
South polar cap
The nadir receives the same projection-aware treatment
-
4
Smooth recombination
The cap bands blend back into the wrap-aware full frame
Use cases
Skyboxes
For your virtual world
VR Experiences
High resolution for virtual reality immersion
Concept Art
Develop rich high resolution worlds and consistent worlds for your IP project development
Vibe Coding
Use Hextile to agentically generate worlds for your vibe coded video game
Pipeline Integration
Upscale fits into your workflow as the final step after AI diffusion processing, or apply it directly to an existing panorama. Open the result and use Resize → Upscale; use Resize's Size section instead when the delivery needs an absolute dimension.
Technical Specifications
| Model | Real-ESRGAN (NCNN implementation) |
| Scale Factors | 2× / 4× / 8× / 16× targets, up to the 16K edge cap |
| Input Formats | PNG, JPEG, TIFF, EXR |
| Output Formats | PNG (16-bit), JPEG, TIFF |
| Max Output Resolution | 8192×4096 native, up to 16384×8192 with Real-ESRGAN upscaling |
| GPU Acceleration | NVIDIA CUDA / Vulkan |
| VRAM Usage | ~2-4 GB (tile-based processing) |
| Availability | All licenses |
Work on your own GPU
One-time licence from $360. No subscription, no per-render cloud charge.
30-day money-back guarantee on every purchase.