Shapezo and Meshy can occupy different layers of an AEC or visualization pipeline. Shapezo organizes architectural intent in a complete site context. Meshy generates candidate 3D assets from prompts or references. The integration question is not which preview looks better. It is which system owns the next decision.
Keep model states explicit
An AI-generated mesh is an exploratory artifact until it passes checks for scale, orientation, topology, materials, and downstream use. A Shapezo option is also provisional, but its value is the relationship between envelope, public edge, circulation, landscape, and neighboring American buildings.
Use explicit states such as exploratory_asset, concept_scene, reviewed_visual, and approved_for_visualization. Avoid a single final flag that hides uncertainty. This prevents a plausible Meshy object from being interpreted as verified architecture after import.
Put context around generated geometry
Meshy is naturally object-centric: a prompt describes a thing. Architectural review is context-centric: a canopy matters because of its effect on an entry, a sidewalk, a view, or a shadow. Shapezo can provide the scene in which those effects become testable.
Store the generated asset ID, source prompt, reference hash, units, up axis, bounding box, and asset version. Link those records to the Shapezo option that consumed the object. The relationship is more valuable than a one-time export.
A minimal exchange contract
For a building-led workflow, a compact payload can include:
-
project_idandoption_idfor traceability. -
asset_idandsource_typefor prompt, reference, or manual origin. -
units,up_axis, andcoordinate_referencefor predictable placement. -
bboxandlod_intentfor scale and performance expectations. -
validation_statesuch as exploratory, reviewed, or approved-for-visualization. -
ownerandnext_checkso unresolved issues have a destination.
The exact schema can vary. Geometry should still travel with enough metadata to explain what it is and what it is not.
Review gates that save rework
Generate several candidates in Meshy, normalize names and textures, and place only the strongest options in Shapezo. Review street scale, shadow behavior, visual weight, access, and relationships to the surrounding scene. Reject assets early when they create a problem that an isolated preview concealed.
Performance and reproducibility
Generated assets can vary in polygon count and texture weight. Create a lightweight review derivative while retaining the original output for provenance. Cache the prompt, reference, generation time, and scene version so a selected result can be located again.
Keep a Shapezo presentation snapshot separate from an editable option. When an asset changes, record the reason and the receiving scene version. This turns visual iteration into an inspectable event instead of an unexplained overwrite.
Decision rule
Use Meshy when the system needs more candidate geometry. Use Shapezo when the team needs to understand what candidate geometry does in a place. Connect them with explicit state, metadata, and human review before either becomes a project commitment.



Top comments (0)