3D printable object extraction

[Dune]

I wonder if anyone knows the answer (or is here to read my query), but nevertheless a try. I was asked to produce a 3D printable object from parts of a few of my scenes. Has anyone done this? I suspect an animation (without blur) surrounding a scene would yield renders that could be used to import in a photoscan software, but the practise is often more difficult than the idea...

[gao_jian11]

I tried to render a series of images in a loop in tg and then use RealityCapture to generate the model, but it's time consuming, if it's for 3D printing, I think it would be faster to export the mesh model directly in tg and then add the plant and house models externally. The accuracy is also geng higher.

[Hannes]

Hmmm, I have no idea, but exporting parts of a TG scene to make a real 3d model for making a clip that looks like the initial TG scene animated sounds somehow weird (if I understood it correctly...).

[Dune]

Thanks for your input guys. I've just downloaded some free photogrammetry apps, and might have a go at it. It doesn't need to be really hires, I guess, as it also needs to be printed in a robust way. So no grasses, but a humpy ground. The characters and larger bumps in the ground should however be good enough.
With time consuming, you mean the actual loop render or the photogrammetry work? I think I would start off with a lowres render. No lights or clouds are needed so that speeds up rendering. Just GI maybe, or some lights without shadow surrounding the scene?
Direct export is a bit more work, I guess, as the final model also needs to be watertight. And with photogrammetry taht would probably work better; one model of all stuff in the scene.

[KlausK]

"And with photogrammetry taht would probably work better"
I don`t think so. The meshes generated are very messy most of the time.
That will be a lot of work, I guess, to get the photogrammetry object to be 3d printable.
You will have to remesh / rebuild the model to make it watertight (I think).
Even getting a usable photogrammetry object with free software...

Any idea of how big (live dimensions) this is going to be?
Or is that work the client does himself afterwards?

Is that FDM or SLA (Resin) printing?
You should know ask for with what kind of printer they are going to print.
Also in regards to the printable area the printer can do.
That makes a difference for the model you have to generate.

If you know more, maybe I have a few more tips for you.
Very curious to see what becomes of that project!


CHeers, Klaus

[aknight0]

I did a video about this a few years back, which may give some ideas of the results:  https://www.youtube.com/watch?v=vA4gHBTfKPM&t=1050s&ab_channel=BYUPRISM

Photogrammetry models aren't usually very good for printing, since the meshes produced are very messy.  I'd expect that you'd have better luck with the microexporter, but there will be post-work either way to make the model printable.  

[gao_jian11]

GI also does not need to be on if the model is only being output for 3d printing. 3d printing also does not require colour information. Materials can also use the simplest bit of noise, in order to provide a basis for measurement by photo modelling software, which does not recognise smooth, white surfaces well.

[gao_jian11]

I wonder if the output scene overhang structure is important? If there is no direct output of height maps, generating obj prints in other software works best. I have done CNC engraved output of obj terrain.

[Dune]

Thanks again guys, really appreciate your thinking along.
@KlausK ; I don't know what they will use as material, nor size, but I expect it to be robust stuff (outside, so vandalism-proof) and not too big, maybe 50cm across. It will be a scene of say 10x10 terrain with some dents and square peatholes, and some working characters digging peat, for instance.

I tried making one of my people models (consisting of several parts and layers) into a watertight printable object, but even that isn't easy. ZBrush does a pretty good job at making all the parts into one mesh by dynamesh, but lots of holes still appear. Closing them still gives 'closed holes', like in an overhanging tunic. Remeshing in ZBrush will just give an untextured model, and this project needs color preferably.
 
That also answers your question @gao_jian11 ; overhangs are important for the characters, not so for the terrain. For that I can indeed use the TG microexporter, or through a depth mask and ortho render. Thanks for your video, interesting to see your workflow.

So I wonder how these guys make their scanned treetrunks so nicely closed and colored.... they mostly look good for printing as well. Also, if I were to make separate models for characters and terrain, they will have to be merged for a one go print, I guess. More digging to do....

[Hetzen]

I think you'll be looking at printing this in individual parts. Not only do you need a closed mesh with no holes or intersecting geometry (which creates inverted volumes), you'll be getting into supporting the geo with supports for either filament or resin printing.
There is sinter printing, which is very expensive, especially a 50cm cubed object. This method doesn't need supports though and doesn't mind over hangs or 'island' geometry.

I'd be thinking of printing your base terrain ground, then shrubbery and figures as separate objects, then paint and glue together. Woodland Scenics do some great static grass and clump foliage to help bring this all together.

[Dune]

Hey Jon, good to hear from you. Thanks for the input! I'll put this to the client as well. I am pretty curious how deep their pockets are