Quickly take an AI generated image and make it a 3d print with the following steps.
- Generate an image
- Make the image a 3d model
- Translate the model to something 3d printer software understands
- Print (after coloring the model optional)
Generate an image
Use Flux or some other image generator to create an image. Be sure to include something like "Video game concept art pure white background" or "bright green background" to have the ability to remove the background easily.
Use Convert the image to a 3d model
I used TRELLIS to generate a 3d model from my image.
Translate the model to something a 3d printer can understand
Most slicer software needs stl or 3mf (although many can import from other file types). You can use a tool like Image-to-STL or open in Blender and export as STL/3mf.
Print the Model
Now you should be ready to print your model. I re-colored it in Bambu Studio, as I wasn't able to figure out a way to keep the colors intact through the filetype translation process.
And the result
Let me know what you think.
Top comments (5)
I am very curious about how this space will evolve. It's definitely an interesting pipeline.
Yea! I think even just having the ability to import the colors would go a long way BUT I think we'll have a day in the near future where we can prototype machinery and stuff.
I imagine it makes sense to also have a 3D-object-native AI generation technology to directly generate the actual 3d files natively with the intelligence baked into that process vs first doing an image then generating to 3d — or at least that makes sense to me. But you understand this area more than me (I'm a newbie with 3d printing), do you think that's where it will go?
Not sure that I would necessarily understand more than you, however, I do wonder if we're a little ways off before we would have the ability to say "Make this thing that has many moving pieces" I think there would need to be almost like a LangChain type thing for orchestrating the various parts and establishing the types of connections that are made like "This type of bolt goes here to hold __ and __ together." That said, as fast as this is moving in other areas I wouldn't be surprised if this is already solved or solved soon (BUT I'm not an expert at that in any sense)
I am a Bamboozled X1C owner and operator myself. I think the results Ryan shows leave a lot to be said about fidelity of the finished model. Hopefully we'll see compute improve this over the next year so unquestionable quality can be generated.