From SketchUp to Unreal: a 2026 stack for archviz studios
What we run, why we run it, and which combinations we'd avoid in 2026.
There's no single 'right' archviz stack — there's the stack that fits your projects, your team, and your delivery model. Here's ours, broken down by phase.
Modelling — SketchUp, Rhino, Blender
Architects send Revit and SketchUp. Designers send Rhino. Developers send the back of an envelope. We pull all three into Blender for unified shading and re-light. Blender's modifier stack is what makes a 32-storey tower modellable in days, not weeks.
Lighting & rendering — Corona, V-Ray, Cycles
Corona for interiors (best out-of-the-box realism). V-Ray for exteriors and animations (predictable at scale). Cycles in Blender for product work and any project that doesn't need 3ds Max. The right answer depends on the shot — that's why we don't have a single house engine.
Real-time — Unreal Engine 5
Anything that needs to run interactively (VR walkthroughs, training, configurators) goes through Unreal. Lumen and Nanite have closed the gap with offline rendering for 90% of archviz use cases. The remaining 10% is hero shots that still belong in V-Ray.
AI layer — Stable Diffusion XL, ComfyUI
Brand-fine-tuned SDXL for variation passes, orchestrated through ComfyUI. We run our own GPUs because we won't put client IP through a third-party API. This is the layer that's changed most in the last 18 months — what we run today won't be what we run in 2027.
What we'd avoid
Rendering anything in Cinema 4D for archviz (it's a great motion-graphics tool, not an archviz tool). Using purely AI-only generators for client deliverables. Trusting any cloud rendering farm with NDA work without an explicit data-handling agreement.
