Cityweft vs Shapezo: A Developer's Guide to Map-Driven and Scene-Driven 3D City Models

I wanted to compare Cityweft and Shapezo as workflows, not as feature checklists. The useful question for me is where the first geometry comes from and how much control I need after that first result. Cityweft lets me assemble and edit an urban scene. Shapezo lets me select a map region and ask AI to generate a model for it.

1. Define the Input Contract

For Cityweft, my input is a scene plan: buildings, roads, public spaces, terrain, and environmental context that I want to organize together. I decide what belongs in the scene and how the parts relate. The workflow is explicit, which makes it easier to reproduce a design study.
For Shapezo, the key input is geographic scope. I draw a rectangle on the map, and the AI uses that selected area as the modeling boundary. The box is a lightweight contract: include this region, interpret its spatial structure, and return a 3D starting point.
The first engineering lesson is that these tools do not solve the same input problem. One takes a composition; the other takes an extent.

2. Inspect the First Result Like a Prototype

I treat both outputs as prototypes. I check whether the terrain has the right broad shape, whether buildings sit on plausible surfaces, and whether roads connect in ways that make sense. A model can look smooth while containing a bad assumption about elevation or access.
This is where I use a simple review loop:

1. Rotate the model through ground, oblique, and top views.
2. Compare major edges with the source map.
3. Mark uncertain geometry instead of silently accepting it.
4. Regenerate or edit only the parts that affect the design question. The loop keeps the review focused. I do not need to validate every facade when my actual question is whether a corridor can connect two districts. ## 3. Choose the Right Post-Generation Control Cityweft is useful when I need to modify the model as a design system. I can change building massing, street organization, and public-space relationships while keeping the scene legible. That is valuable for option studies because I can compare a baseline with a deliberate alternative. Shapezo is useful when the unknown is the place itself. The map selection gives me a fast regional draft, which I can use to identify the parts that deserve more careful modeling. It reduces the cost of the first orientation step, especially when the site is unfamiliar.

4. Keep Provenance With the Model

I save the map extent, source date, coordinate assumptions, and generation settings alongside screenshots. For Cityweft studies, I also save the version of the scene and the changes made between options. For Shapezo studies, I record the exact rectangle because a small boundary change can produce a different context.
This is not bureaucracy. It is how I keep an attractive image from becoming an untraceable claim. When someone asks why a road or block appears in the model, I can point to the source and the decision that produced it.

Where Each Workflow Fits

I use Shapezo early, when I need a broad spatial hypothesis from a real map area. I use Cityweft later, when I want to test alternatives inside a controlled scene. If I combine them, I keep the boundary between discovery and design visible: the generated context is evidence for a question, while the edited scene is a proposal to discuss.
Neither workflow replaces a survey, a GIS database, or engineering documentation. Both can make those later steps easier to explain because the spatial relationships are visible before the technical package is complete. That is the practical difference I keep coming back to: Shapezo helps me initialize a place, and Cityweft helps me iterate on a place.