Letting rendering carry the garment

Rendering can substantially change the impression of a virtual garment. Simple tailoring is especially vulnerable: weak lighting or material presentation can make it look cheaper than intended. I wanted to close that gap.

At the time, fashion teams using CLO3D and VStitcher were familiar with rendering such as V-Ray. To build a service generating high-quality 4K turntables beyond real-time presentation, I first compared three renderers.

My selection notes at the time

These assessments reflect the environment and requirements I was considering for this service.

Comparison under my selection conditions at the time; the same entries follow as text.
Comparison under my selection conditions at the time; the same entries follow as text. — Illustration: Enzoviz — historical renderer selection notes

Cycles

  • Implementation/upgrades: easy
  • Render time: medium (sample 1024)
  • Devices: GPU/CPU
  • Shader compatibility: good
  • File compatibility: blend/fbx/glb/c4d
  • Cost: none
  • Quality: high
  • Farm support: many options

V-Ray

  • Implementation/upgrades: easy
  • Render time: long (4 sub div)
  • Devices: GPU/CPU
  • Shader compatibility: good
  • File compatibility: fbx/glb and a broad range of formats
  • Cost: high
  • Quality: high
  • Farm support: many options

Unreal

  • Implementation/upgrades: complex
  • Render time: short
  • Devices: GPU
  • Shader compatibility: good; separate Blueprint preparation
  • File compatibility: fbx only in this setup
  • Cost: case-dependent
  • Quality: high
  • Farm support: few options

Why I chose Cycles

I ultimately chose Cycles. With Blender 2.8 onward offering a capable feature set and Python API, it fit the pipeline I wanted: assemble scenes in a worker environment, then send them to a render farm.

Soft shadows and the bounce of global illumination let the fabric’s character emerge naturally. I find that interplay of light satisfying just to look at.