I've been getting my first test model ready to print over the last few
days and have encountered numerous problems. I was expecting a certain
amount of problems and fixes that needed doing but I wasn't prepared for
quite so much going wrong!
Firstly I managed to get my test sculpt completely finished, I was actually really happy with it. I decided I didn't want to do the whole model as this would increase the time it took me to make and the cost of getting it printed. So I just did a bust and I got it done from start to finish in about a week, which was pretty good.
Here is some pics of the finished bust.
I even made a little stand for it so it could stand up alone (the base isn't flat).
Despite Zbrush having a 3D printing export plugin, I wanted to export it
into 3Ds Max first so I could fine tune any sections of the model and
rescale anything I wanted easily. Upon importing the .obj file into Max I
was confronted with an import of almost 9million polygons. After
reading some guidelines on the Shapeways website I found out this was
far to complicated. I made the mistake at the beginning of the project
in thinking 'its not for use in a game I can make it as complex as I
want!'. this simply isn't true. The triangle limit for 3D printing (at
Shapeways at least) is 1million. I could work around this by breaking up
my models in the future but for the purpose of the test print I just
needed to simplify it at bit.
This wasn't a difficult process to do within Zbrush. Although I had decimated the model somewhat, believing I had an unlimited budget meant I simply hadn't done it enough. I was very quickly able to get the polygon count of the model down from 9million to 340,000 without losing any real detail from my model. These images show how zbrush likes to decimate a model. (Anyone that knows 3D will know this is a horrible use of polygons, but thankfully the use isn't important, just how many there are!)
This wasn't a difficult process to do within Zbrush. Although I had decimated the model somewhat, believing I had an unlimited budget meant I simply hadn't done it enough. I was very quickly able to get the polygon count of the model down from 9million to 340,000 without losing any real detail from my model. These images show how zbrush likes to decimate a model. (Anyone that knows 3D will know this is a horrible use of polygons, but thankfully the use isn't important, just how many there are!)
After sorting the triangle count down to well within budget I followed more instructions on the Shapeways website (found here) and downloaded a program called netfabb whose website can be found here!
Netfabb is a program primarily used for checking models for any errors or issues that would cause faults when 3D printing. It automatically checks and fixes several issues including; holes, flipped faces, manifold edges and if a model is water tight.
The first issue I faced was due to scaling problems, a model made in Zbrush has completely irrelevant dimensions when put into a program specifying something for real world printing. During fixing smaller parts of my model would disappear. I worked out this was because part of the auto fixing script removed 'shells' (equivalent to elements in 3DsMax) smaller than 0.02mm. Which obviously lead me to question what the overall size of my model was, turns out it was really really tiny. My model in netfabb.
Netfabb is a program primarily used for checking models for any errors or issues that would cause faults when 3D printing. It automatically checks and fixes several issues including; holes, flipped faces, manifold edges and if a model is water tight.
The first issue I faced was due to scaling problems, a model made in Zbrush has completely irrelevant dimensions when put into a program specifying something for real world printing. During fixing smaller parts of my model would disappear. I worked out this was because part of the auto fixing script removed 'shells' (equivalent to elements in 3DsMax) smaller than 0.02mm. Which obviously lead me to question what the overall size of my model was, turns out it was really really tiny. My model in netfabb.
Netfabb
has tools in it to re-scale models, which auto scale all the dimensions
to retain shape. The process of rescaling was therefore pretty simple,
but it lead to another problem, 'how big did I want the model?'
This was a little harder to fix as I found there needs to be a balance between size and affordability. The purpose of the test print was to see how much detail I could get into a model that got progressively smaller and therefore finding the point at which a model was too small to get in the detail I wanted. Getting a big 4inch model printed (although awesome) would have been pointless to my experiments. I measured various GamesWorkshop models that I own to get a general idea of how big I wanted my model to be.
The first model I uploaded to Shapeways was a modest 2.5inches. I thought this would have been a good compromise, it was bigger than most Games Workshop models (apart from Forge World specials) but not so big it was pointless to do.
Shapeways has an automatic filter which checks the suitability and details of models uploaded. I had made sure my model fit all these specifications before uploading, apart from one, which is different depending on which material you're printing in. The wall thickness. This lead to another more major problem with my model.
My model uploaded to shapeways with details, prices and warning.
This was a little harder to fix as I found there needs to be a balance between size and affordability. The purpose of the test print was to see how much detail I could get into a model that got progressively smaller and therefore finding the point at which a model was too small to get in the detail I wanted. Getting a big 4inch model printed (although awesome) would have been pointless to my experiments. I measured various GamesWorkshop models that I own to get a general idea of how big I wanted my model to be.
The first model I uploaded to Shapeways was a modest 2.5inches. I thought this would have been a good compromise, it was bigger than most Games Workshop models (apart from Forge World specials) but not so big it was pointless to do.
Shapeways has an automatic filter which checks the suitability and details of models uploaded. I had made sure my model fit all these specifications before uploading, apart from one, which is different depending on which material you're printing in. The wall thickness. This lead to another more major problem with my model.
My model uploaded to shapeways with details, prices and warning.
The material I've chosen to
get my model printed in is a strong & flexible white plastic that is capable of
printing high levels of detail. This material (and most others) requires
that no section of the model is under 0.7mm thick as it will be too
thin and could snap. So after uploading I got a warning saying my model
had 'very thin walls' meaning some sections of my model where bellow the
0.7mm required.
I took my model back into netfabb, which has a handy measuring tool allowing you to quickly see and identify parts of the model that are too thin. Unfortunately it was practically every single detail on my model. My first reaction was to just make the model bigger until all the parts where thick enough, I found this made the model cost jump from $33 to a staggering $170, and yet I had only doubled it to 5inchs. I obviously couldn't spend that amount of money on a test model so I had to start rescaling parts of the model.
During the process of rescaling parts I became increasingly frustrated! The model looked fine in zbrush and max, everything looked in proportion and accurate, so why where the details becoming so small when the model was still relatively large at 2.5inches. I'm still not even sure. I reevaluated all the games workshop models I have and went through some carefully measuring the thickness of the smaller details. Although there are some very small fine details, no matter how small the games workshop models get, the details are always approximately 1mm or bigger. So I realized my model is just too complicated, it has too many details and they are far to small. Everything needs to be simpler, bigger and chunkier.
Rescaling a model is extremely difficult and inaccurate when the model has been finished and the geometry is really poorly laid out (thanks to decimating in zbrush). I had to make a choice whether to take a few steps back, return to zbrush and completely redo large sections of my model, continue to attempt rescaling or just payout for a massive model. Given this was just a test model I didn't want to put any more time into it or spend too much money. So I continued attempting to fumble around with rescaling. Another idea I had was to get rid of the base, although it looked nice and made the model look like a tasty bust to sit on my desk, it wasn't needed at this time. Removing the stand detail and shortening it meant that even with the model still at 2.5inches more of it would now be the main model and not the stand, meaning the chest is closer to 2inches than it was before.
Rescaled model with less base!
I took my model back into netfabb, which has a handy measuring tool allowing you to quickly see and identify parts of the model that are too thin. Unfortunately it was practically every single detail on my model. My first reaction was to just make the model bigger until all the parts where thick enough, I found this made the model cost jump from $33 to a staggering $170, and yet I had only doubled it to 5inchs. I obviously couldn't spend that amount of money on a test model so I had to start rescaling parts of the model.
During the process of rescaling parts I became increasingly frustrated! The model looked fine in zbrush and max, everything looked in proportion and accurate, so why where the details becoming so small when the model was still relatively large at 2.5inches. I'm still not even sure. I reevaluated all the games workshop models I have and went through some carefully measuring the thickness of the smaller details. Although there are some very small fine details, no matter how small the games workshop models get, the details are always approximately 1mm or bigger. So I realized my model is just too complicated, it has too many details and they are far to small. Everything needs to be simpler, bigger and chunkier.
Rescaling a model is extremely difficult and inaccurate when the model has been finished and the geometry is really poorly laid out (thanks to decimating in zbrush). I had to make a choice whether to take a few steps back, return to zbrush and completely redo large sections of my model, continue to attempt rescaling or just payout for a massive model. Given this was just a test model I didn't want to put any more time into it or spend too much money. So I continued attempting to fumble around with rescaling. Another idea I had was to get rid of the base, although it looked nice and made the model look like a tasty bust to sit on my desk, it wasn't needed at this time. Removing the stand detail and shortening it meant that even with the model still at 2.5inches more of it would now be the main model and not the stand, meaning the chest is closer to 2inches than it was before.
Rescaled model with less base!
I eventually got to a point where the thinnest parts of the model had been bulked out, some still didn't meet the 0.7mm guidelines but I re-uploaded it anyway. I got the warnings about wall thickness but ignored them, a technical guy checks over the model if this warning pops up after purchase, so I thought he could deal with it!
The new price of my model at 2.5inches is $60, which I have guessed is due to the model now having a higher % of detail as I have shortened the stand but kept the size the same. This is still more money than I wanted to pay, but I cannot make the model any smaller than it currently is. As this model is so much larger than I am likely to have any future ones printed I am no longer experimenting in the level of detail a small model can achieve, so I briefly thought why am I even bothering to get this printed at all?
I then realized there is still plenty of stuff I can learn by getting this test print done. I can see how long it takes to get the model checked, printed and shipped, I can also see how long shipping from overseas will add to this time and if the models are likely to get damaged in transit. I will also be able to see upfront what the material is like, how good the printing is and what details at various sizes actually come out like.
So I went ahead and purchased the model, only to have shapeways email me a few hours later saying they can't print it and gave me store credit back. Thankfully shapeways didn't just say 'no', within the email they included the reason, which turned out to be sections of the model where floating freely from the rest, which would obviously fall if printed.
Here is the screenshot they sent me back, highlighting the issues.
I looked back at my model and during all the rescaling (remember I said it was hard once it had all been built), parts had moved away just as shapeways had said. These were quick and easy to fix so I had the updated model re-uploaded within an hour or two. I purchased the model again using store credit and am currently waiting to see if shapeways confirms or rejects the model for printing.
Awesome Luke. You are really pushing the boundaries here and showing great dedication.
ReplyDeleteThanks very much! I'm looking forward to getting the model back and moving forward from there. (Provided they print it)
Delete