Full Version: Ai export optimisation

Show messages: All  1-2  3-16

Hi Matt,

I agree it would be nice to have less segmented outputs.

You must probably know this, but just in case :

The colours of the object's layer are exported so it's relatively easy to select all lines of the same color in Illustrator and put them in a new layer.
Also, the color of different additional lines can be set in the options (silhouette, hidden. etc..) .


Marc

---

Hi Matt, I am working on answering your message, it's just taking me a while to gather things up.

- Michael

---

Hi OStexo

Yes I usually use illustrators simplify path command with accuracy set to 100% to reduce curves with 100s of points down to a fraction of this.

Matt

---

Hi Matt, well I do have to mention that it's kind of a stretch to expect a $300 program made by one person to do the same job as a $10,000 program from a big corporation - just sayin' ... ;) However, I would like to improve it to work better for you.

> Firstly, and this would make a huge difference. Is there any way to limit an ellipse to the four
> points on major and minimum axis instead of random scattering. See attached for explanation.
> Maybe an extra check box "export true circles as ellipse"

I think that's pretty feasible. The part that makes it a bit complicated is that kind of four point circle or ellipse in AI is not actually an exact circle or ellipse, it's an approximation. It's not possible to represent a circle or ellipse exactly with a simple 4 point cubic Bezier curve, doing general conic sections requires what's called a "rational" curve (where the R in NURBS comes from). But AI does not support rational curves, so any rational or high degree curve in MoI will get converted by a generic fitting process into a non-rational cubic curve that's then what is exported. This fitting process rebuilds the curve by sampling points along it and making a new curve through those points, adding more points in between the current ones until the new curve is within a tolerance value of the old one. It's difficult to make this process as good as someone using human judgement to place points. But handling ellipses as a special case should be possible.


> Secondly, when selecting to have the silhouette lines as well as standard lines, the
> export gives double lines, one for each. it would be much better if it was a single line
> as in the attached. It becomes very problematic when cleaning up the file as you find
> there are many lines on top of each other.

This is probably not too difficult, I can look into it some more after the next beta is out.


> Thirdly, I notice on large size (dimension not mb) models especially that MOI brings in
> some of the linework which is obviously hidden.

It tends to be just generally difficult when there are pieces of different scales involved, tolerance values that would be appropriate for the large piece are too large for the small piece.

re: Canvas size - currently that is limited to a max of 175, but now with 64-bit I'll raise that limit to 1000 instead. Bumping this up consumes quite a lot of memory though, it takes up about 34 MB per megapixel and putting it at 1000 will consume about 34 GB and may not be practical but you'll be able to give it a try. An increased density should help with better generation of small areas.

Typically when you have a model like this with a mix of small and large pieces, are the small pieces usually separate objects and not just some small feature of one big solid object? I have an idea that maybe I can improve things by doing some initial work on each object individually (with the canvas tightened up to just surround that object) to determine its own self occlusion areas before then doing a final pass with everything combined.


> Lastly, is it possible to translate MOI layers into illustrator layers or groups on
> ai export. this would make selection and work much easier.

Yes, this is possible currently - by default MoI will write layers to the .ai file named "Silhouettes", "Visible lines", "Background image", etc... but you can tell MoI to use the generating object's style in MoI instead of that. To set that up, on the main "AI export options" dialog click the "Line style options" button, and then on the Line style options, change the Layer option from "Custom" to "From style":







- Michael

Image Attachments:
ai_layers1.jpg  ai_layers2.jpg  ai_layers3.jpg 

---

Hi Micheal

Thanks for the detailed answer. Yes I appreciate that its just you which makes this software even more impressive :-) The 10k software was a revelation when it first came out which probably justified the price tag (it also put about 80% of tech illustrator out of business . . . ). It was also developed by a small team for the Mac first, but as it got older it never really progressed. It is in my opinion a one trick pony, if i want to add a part, or modify then I had to go back to the CAD user to get a new STP file. This is why I prefer working in MOI - If i need to show aircraft door in open state i simply place in its own group and rotate about hinge axis - easy. In the other software this could take me ages, for a start i could re arrange parts into separate groups so if the file wasn't properly organised I had a real selection headache ahead of me. Also even now MOI gives a cleaner output in many cases.

Enough of the past, MOI is where its at for me :-)

So to answer your question "Typically when you have a model like this with a mix of small and large pieces, are the small pieces usually separate objects and not just some small feature of one big solid object?"

If it is a small piece, say a wheel on aircraft then yes I can export just the wheel and it will be very accurate. My problem is when I export say the hydraulic system - its made up of small pieces but spread along the full extent of the aircraft.

Regarding the cleaner ellipse with 4 points - even standard 8 points would be good. One of the main points is to have points at the outer extent of the ellipse so you don't have to cut at this point when making thick and thin, the point is already there.

Thanks for your work on this, I fully understand that this is a niche area and your time is limited, any improvements are always very much appreciated.
BTW below is the kind of thing I do. My next project, another aircraft, I want to do using MOI only for the ai export. I willl let you know how it goes :-) Image created in Adobe Illustrator.

Matt

Image Attachments:
Screen Shot 2017-12-07 at 08.52.11.png 

---

Holy guacamole, that is an impressive amount of detail. Do you ever sleep? :)

---

Hi Matt, wow that's a gorgeous plane illustration!

re:
> If it is a small piece, say a wheel on aircraft then yes I can export just the wheel and it will be
> very accurate. My problem is when I export say the hydraulic system - its made up of small
> pieces but spread along the full extent of the aircraft.

So what I was getting at there is not so much that you should export just the wheel, but rather that I could probably make the hidden line exporter do an initial pass that worked on objects individually with the whole canvas temporarily framed around just that object. That could then give objects better accuracy on an individual level and the limits of the canvas resolution being stretched across the full scene would only impact how objects interacted with other objects and not also gunk up a small object's own self occlusion.

It may be a fair chunk slower than the current exporter since it would be doing more work for each individual object but would it be accurate to say that even if the exporter was say 5-10 times slower than current it would save you time in the long run if it gave cleaner results on small objects within a larger overall scene?

- Michael

---

An idea of time drawing this plane ?

---

Thats sounds like a good workaround. I assume the canvas size is calculated on what is visible?

I don't mind the speed hit, its always better than having to redraw things which didn't quite come out right. At the moment I barely have time to make a coffee let alone drink it when i export :-)

matt

---

this illustration took about 3 weeks, big chunk of that time is searching for all the 3d parts in NX and exporting.

In the old days to do an airbrush job would be a bout 4 weeks but this is 100% accurate, down to the last rivet if I wanted, although i usually strip out a lot of fixings in something this big. And the bonus is I can add extra bits if need - just remember the viewing angles and export at same size and position and it slots right in.

matt

---

That's an amazing illustration Matt !

Marc

---

Thx for the infos!

---

Hi Matt,

> I assume the canvas size is calculated on what is visible?

Yes, it covers the same area as the shaded background image that you optionally generate on AI and PDF output.

I've got these improvements on my list now, I should be able to work on them once the current V4 bug list is cut down.

- Michael

---

Brilliant! :-)

---

Show messages: All  1-2  3-16