I used an AI generated costume concept as the starting point for a set of crack shaped pieces, then turned a simplified image into a 3D model with Tripo. The important work happened after generation: I flattened the model in the slicer, cut it down to two layers, and printed several sizes in TPU. The samples bend around my hand, which is the behavior I wanted for a costume that moves with its wearer.
This is a walkthrough of the pieces shown in my video. It demonstrates the image, model, slicing, and sample prints. The finished decorated costume and its attachment method are future work, so the samples do not establish how well the pieces stay attached during wear.
The result and the key decision
The video shows several black TPU cracks in different lengths, including a smaller piece that I curve around my hand. There is also a decorated sample. Adding stones makes that sample less flexible, although it still curves. My plan is to use multiple crack shapes across the costume and stone them so they read more clearly on camera.
The key decision was to make a thin, mostly flat decoration instead of treating Tripo's full generated volume as the finished part. A dimensional model looked promising in Tripo, but its top and underside needed to be cut before it would serve as a flexible costume element. That distinction is useful for other image-to-3D projects: generation supplies a shape; the slicer supplies the physical thickness and print plan.
Start with a single crack image
I began with a broader AI generated costume concept. It helped me decide where the crack effect could go and what the design should look like. I then generated a separate image of one crack on a clear or transparent background. That simpler image was the input for Tripo. It gave the generator one subject to interpret instead of asking it to reconstruct the whole costume or the body underneath it.
The image came through a phone based workflow, but the important input property is the clean, isolated crack silhouette. A busy scene would leave more background for the model to guess at. For another costume design, start with one element that can stand alone as a print and keep the intended outline easy to see.
The isolated crack design. Source: finished video, 0:59.
Generate a mostly flat model in Tripo
In Tripo I went to my workspace, chose Image, selected the crack image, and added a description asking for a flat background and a printable file. I generated multiple image options and compared them. Some looked too dimensional for the part I wanted; I picked the version that appeared to have a mostly flat top and bottom.
I did not use multiview generation for this project. The intended object was a flat costume accent, so a detailed reconstruction of unseen sides was less useful than a shape I could lay on the build plate and thin down. That is specific to this decorative part. If the back or fit of another generated object matters, check or provide those views rather than assuming this single image route will capture them.
After generating the 3D model, I tried different results and chose one that was close to flat. It was not perfectly flat out of Tripo, and I expected to finish that job in the slicer. I exported the model as a 3MF and opened it for print preparation.
The generated form is a starting shape; the final thin profile came from the slicer. Source: finished video, 2:18.
Flatten and thin the model in the slicer
I prepared the model in the FlashForge slicer and planned to print on my Creator 5 Pro, which already had TPU loaded. In the video I orient the imported crack on one side, adjust its size, and use the slicer's cut tool. I first cut away the uneven bottom, flip the remaining shape, and cut again to get a flat piece. Then I cut the top to set its final thickness.
For this sample, I wanted two printed layers. The first layer was set to 0.25 mm and the additional layer to 0.20 mm, for a target thickness of 0.45 mm. I had already tried different thicknesses to judge stiffness, and this was the thin version I chose for the demonstration. It is a recorded choice for these samples, not a universal TPU setting. A different material, nozzle, profile, or costume requirement may need a thicker part.
I used the cut height controls to reach that thickness, then sliced the model and looked at the preview to confirm that it actually contained a first and second layer. This preview check matters more than trusting the cut dialog alone. If the preview shows missing branches, disconnected islands, or extra layers, revisit the cut position and scale before printing.
The second cut controls the final thickness. Source: finished video, 5:44.
A practical slicing checklist
- Lay the generated shape on the intended print face and inspect its underside.
- Cut away uneven geometry so the part sits flat.
- Flip or reorient it and set the thickness needed for the design.
- Slice and count the layers in preview; check that narrow branches remain printable.
- Set the final length and duplicate only the sizes that fit the intended placement.
- Confirm the printer, TPU profile, and loaded filament before sending the plate.
For broader checks on generated geometry, see my AI generated STL printability guide. It covers scale, fragile details, supports, and slicer review beyond this flat accent example.
Print multiple sizes in TPU
Once the shape was thin enough, I copied it and made different sizes for the costume. In the walkthrough I mention roughly 6 inch, 8 inch, and up to about 12 inch lengths, then place as many as practical on a build plate. These are the sizes I discussed for this design, not fixed dimensions that every costume needs.
I selected black TPU, mapped the plate to the loaded filament, and printed the pieces. The finished small sample bends around my hand, demonstrating the flexibility I was after. TPU was a material choice tied to movement: a rigid accent would be less able to follow the body's curve. The video does not provide a full material specification, attachment test, or long wear test. Use the material advisor if you are deciding whether TPU fits your own decoration or another functional requirement.
The small printed sample curves around a hand. Source: finished video, 7:16–7:24.
What changed after adding stones?
The decorated piece in the video is an early example of the planned finish. I point out that the stones reduce flexibility, though the piece can still curve. That is a useful tradeoff: a finish that makes the crack stand out on camera also changes the mechanical behavior that motivated TPU in the first place.
The decorated sample is more visible, but less flexible than the plain TPU print. Source: finished video, 7:39.
Before decorating an entire costume, make a representative sample at the intended length and bend it to the curve it will have when worn. Check the finished decorated piece, not just the undecorated print. The video shows a hand bend, but it does not show the pieces attached to a worn costume, repeated movement, washing, or long term durability.
Is this workflow right for your project?
| If you need... | What this project shows | Check before committing |
|---|---|---|
| A flexible surface accent | Two layer TPU crack samples that curve around a hand | Whether the final decoration and attachment preserve enough movement |
| Several related lengths | The same generated form resized and arranged on a plate | Branch width and preview layers at each new size |
| A precise fit to a body or prop | A visual concept and sample pieces | Take measurements and test placement; the video does not validate fit |
| A fully finished costume | A plan to use and stone multiple cracks | Attachment, comfort, and movement remain to be tested |
If you want to apply Tripo to a more dimensional object, my AI render to 3D print workflow covers the added checks for multiple views, scale, and supports. The Tripo AI to 3D hub collects that guide and further print tests.
FAQ
Did Tripo produce a print ready, flat costume piece on its own?
No. I chose a mostly flat generated result, exported a 3MF, and used slicer cuts to make the print flat and thin. The sliced preview confirmed the two layer result.
Why print the cracks in TPU?
I wanted the accents to curve with the person wearing the costume. The small finished sample bends around my hand in the video. Its flexibility after attachment and long wear has not been demonstrated.
How thick were the samples?
The demonstrated slicer target was 0.45 mm: a 0.25 mm first layer plus one 0.20 mm layer. I had tried different thicknesses before choosing that version for this example.
Did the decorated version bend the same way?
No. I observed that adding stones reduced flexibility, although the decorated sample could still curve. Test your intended finish before producing all of the costume pieces.
Next step
This project shows how I went from a clean crack image to flexible printed samples. The remaining costume work is placement, attachment, and testing the decorated pieces in motion. For another Tripo project, begin with the Tripo AI to 3D guide hub, then use the printability checklist before committing a full plate.