I get fewer surprises when I separate concept generation from city-model assembly. Cityweft and Shapezo fit into those two stages. Shapezo lets me select an area on a map and generate a rough model with AI. Cityweft is where I can combine AutoCAD content, GIS data, BIM models, terrain, roads, and infrastructure into a larger 3D city environment.
The distinction is practical. A generated model is good at starting a spatial discussion. A connected city model is better at carrying relationships between data sources and disciplines. If I mix those statuses, a polished concept can be mistaken for verified project information.
Stage 1: define a bounded experiment
I start with a clear question and a limited boundary. I might test a station block, a campus edge, an industrial parcel, or a riverfront segment. A bounded question makes the Shapezo result easier to judge. I can compare building massing, street access, open space, and the relationship to visible context without asking AI to invent an entire city.
I save the boundary, the prompt assumptions, and the creation date with the concept. I also note what the map does not contain: survey control, buried utilities, detailed grades, ownership constraints, and discipline requirements. This small bit of metadata prevents the model from gaining authority just because it looks complete.
Stage 2: build the city context
Once an option survives the first review, I move to Cityweft. The exact setup depends on the project, but the pattern is stable. I bring in the available AutoCAD geometry, GIS layers, BIM content, terrain, roads, buildings, water, and infrastructure. Then I organize them into a city-scale view where the selected proposal can be understood beside existing conditions.
The benefit is not only visual. The model creates shared spatial references. A road corridor can be reviewed beside a building envelope. A bridge can be seen with its approaches and river. A public space can be studied with the blocks, landscape, and movement routes around it. That is the kind of context that a single AI massing image cannot provide.
Stage 3: convert rough shapes into meaningful objects
I do not need to replace every generated surface at once. I focus on the geometry that affects the decision. If the question is access, I rebuild the road and its connections. If it is skyline or view, I establish reliable building heights and surrounding blocks. If it is flood risk, I need terrain and water-related data that can support the intended analysis.
This is where a city-modeling workflow earns its keep. Buildings, roads, terrain, and infrastructure can be inspected as parts of the same scene. Export targets such as CityGML, OBJ, STL, PLY, or FBX may serve different downstream tools, but the important step comes before export: knowing what each object represents and how accurate it needs to be.
Stage 4: connect analysis to the model
A 3D city model can support daylight, viewshed, traffic, energy, and environmental studies, but I treat it as input to those studies. I check coordinate systems, scale, terrain quality, object completeness, and the assumptions behind any derived result. A beautiful scene with missing roads or simplified terrain can produce a confident-looking but incomplete answer.
Where each tool is strongest
Shapezo is strongest at the front of the pipeline. It lowers the cost of trying a few spatial options inside a known map boundary. Cityweft is strongest when the project needs an integrated urban context that can combine CAD, GIS, BIM, and infrastructure information for coordination and visualization.
I use Shapezo to decide what is worth building. I use Cityweft to build the context in which that decision can be checked. The handoff is not automatic, and I do not want it to be. It forces me to identify the generated assumptions and replace them with source-backed geometry where the project needs it.
My implementation rule
Keep concept assets separate, keep source data traceable, and make model status visible. With those rules, Cityweft and Shapezo become complementary stages rather than competing claims. One helps me explore a place. The other helps me understand the place as a connected city system.
