For most of art history, the distance between a mental image and a physical form has been measured in skill.
Sculptors spent years developing the hand-eye relationship that let them translate interior vision into clay, stone, or bronze. Industrial designers trained for decades before they could sketch a form in three dimensions with enough accuracy to communicate it to anyone else. The ability to externalize a three-dimensional idea — to pull it out of imagination and make it visible — was itself a discipline, inseparable from the technical mastery required to do it.
That gap is closing in ways that are worth paying attention to.
AI 3D generation has reached a point where artists, designers, and makers without traditional modeling training can now produce three-dimensional forms from descriptions, sketches, and reference images — and do it quickly enough that the tool becomes part of the creative process rather than a final production step.
I've spent the past year building a workflow around three specific tools, and what I've found is that they're most useful not as substitutes for each other, but as distinct instruments suited to different creative moments. Each one corresponds to a different stage in the journey from initial vision to finished form.
---
## The First Stage — Sketching in Three Dimensions
Every creative process begins with the same problem: the idea exists fully realized in your mind, and almost nothing exists outside of it.
Traditional sketching is one answer. But sketching flattens three-dimensional thinking onto a two-dimensional surface, and the translation loses something. You're drawing a representation of a form, not the form itself. The proportions feel right on paper and wrong when built. The relationships between elements collapse in ways that only become visible when the object exists in space.
Formy 3D functions, in practice, like sketching directly in three dimensions. You describe a form in natural language — the overall silhouette, the material quality, the relationship between parts — and receive a fully textured 3D model in minutes. The input can be as rough as a written description or as specific as a reference photograph.
The value at this stage is not precision. It's honesty.
When an idea is described in words, it tends to be described in its most ideal form — the version that exists in imagination, smoothed of all the contradictions and unresolved details. Seeing that idea rendered in three dimensions, even approximately, surfaces the gaps immediately. A proportion that read as elegant in the mind looks heavy in space. A detail that seemed central to the concept turns out to be illegible at scale. Relationships between elements that seemed harmonious in description reveal tensions when they occupy actual dimensions.
I generated multiple form studies for a recent project in a single session with [Formy 3D](https://formy3d.com). Several were wrong in ways I could not have articulated without seeing them. One surprised me — it resolved a tension in the original concept that I hadn't known was there. That kind of discovery is what the sketching phase is supposed to produce, and it came from the same rapid, low-stakes iteration that good sketching enables.
The tool supports multiple input workflows: text-to-3D for concepts that exist only as description, image-to-3D for translating visual references into three-dimensional form, and multi-view generation for objects with significant geometry on all sides. The output formats — FBX, GLB, OBJ, STL — cover the range of what the refinement stage will need.
---
## The Second Stage — Resolving the Form
Sketching produces direction. Refinement produces specificity.
Once a concept has been identified as worth developing, the creative questions change. The work is no longer about finding the right form — it's about resolving that form precisely enough to exist in the physical world, to be communicated to collaborators, or to be tested as a material object.
Copilot 3D is the tool I use at this stage, and the distinction that makes it valuable here is its multi-reference prompting. You can feed it several reference images alongside a text description, and the model synthesizes them toward a more specific result than any single input would produce. This matters when you're working toward a defined creative vision rather than exploring open-ended possibilities.
For artists and designers working toward physical realization, the STL export is significant. STL is the format that connects digital form to physical production — it goes directly to a 3D printer, bringing a designed object into material existence in hours. The output is not always production-perfect without adjustment, but it is close enough to function as a working prototype: something you can hold, test the weight and scale of, examine in actual light, and make decisions from.
There is something that happens when a form that has existed only in digital or mental space becomes a physical object. Decisions that seemed resolved on screen reveal themselves as open questions when the form is in your hand. Materials that seemed appropriate in the render feel wrong in physical scale. Proportions shift in meaning when gravity is introduced. The prototype stage is not just a step toward manufacturing — it is a stage in the creative process, one that produces specific knowledge unavailable by any other means.
Copilot 3D also functions as a communication instrument. Sending a 3D file to a fabricator, a gallery, or a collaborator — rather than a flat render they cannot examine from other angles — changes the quality of the conversation. The recipient can open the file, rotate it, evaluate it from positions the original photographer or renderer did not choose. The form speaks for itself rather than through a curated presentation of it.
---
## The Third Stage — Presenting the Work
There is a meaningful difference between a form that communicates geometry and a form that communicates presence.
The first is sufficient for internal creative decisions. The second is what's required when the work is being presented to an audience — collectors, curators, clients, collaborators whose engagement with the work depends on how it looks, not just what it is.
Trellis 2 is designed for this presentation layer. The distinguishing technical feature is PBR texture support — physically-based rendering — which means the materials on a model respond to light the way real materials do. Metal reflects. Matte surfaces absorb. Fabric carries depth and variation. This is the difference between a render that reads as a digital approximation and one that reads as a document of something that exists — or could exist — in the world.
The workflow at this stage: take the resolved form from the previous stage, generate a presentation-quality version through [Trellis 2](https://trellis-2.net)'s image-to-3D pipeline, and produce multiple material and colorway studies in the same session. For gallery proposals, portfolio materials, or client presentations, having several visually distinct treatments to consider is more useful than a single polished version.
Trellis 2 also handles multi-view input, which matters for objects where the rear, underside, or secondary profiles are part of the design language. A form that has been conceived in the round deserves to be rendered in the round, and the multi-view capability makes that achievable without manual reconstruction.
---
## What This Workflow Actually Changes
The three-stage structure — rapid form exploration, physical refinement, presentation-quality rendering — is not new. Artists and designers have always moved through these phases. What's changed is the cost of each transition.
When generating a form study took days rather than minutes, the creative process was structured around that constraint. You committed to fewer ideas, held onto them longer, and invested more in each iteration before moving to the next. The economics of the process shaped the creative decisions inside it.
When the same generation takes minutes, the creative process can operate differently. You can explore more directions before committing to one. You can test a physical prototype before investing in fabrication. You can produce multiple presentation treatments before selecting the one that best serves the work.
Formy 3D, Copilot 3D, and Trellis 2 don't change what it means to make something considered. They change what you can afford to discover along the way.
---
For artists working in three dimensions — whether in sculpture, product design, ceramics, jewelry, or any form that occupies space — the question worth asking is not whether AI generation is relevant to your practice. It's which moment in your existing process it could make more generative.
Start with the stage where you currently feel the most constrained, and see what changes when that constraint is removed.
