Trouble with filleting

Hi all,

Thought I’d try to model a guitar body. Outline is imported from Illustrator & I am using PlanarSrf and extruding.

However, when I try to fillet, or blend edges, I’m getting errors. It seems to create the fillet, but not connect it to the original object.

Not sure what could be causing this or how to correct. FWIW there are more control points appearing in Rhino than there are in Illustrator, so maybe that has something to do with it?

guitar-body.3dm (247.1 KB)

If you look at your outline under CurvatureGraph you can see a few trouble-spots where the outline isn’t properly smooth. Of course by using FilletSrf instead of FilletEdge it’s usually possible to hack those out, but (gets up on soapbox) that’s the problem with using Illustrator curves in Rhino, they’re not smooth enough for “cutting steel.” Fonts, graphics, assume they’re all garbage and need to be redrawn.

1 Like

I checked it out - it may be a “problem with Illustrator” but I’d like to find out how to remedy it other than simply re-building curves. Avoiding Illustrator is a non-starter.

FWIW I quickly rebuilt it in Rhino and ended up with some of the same nonsense. I then created a shape in Rhino and intentionally tried to keep it really smooth, and still ended up with some bumps in the Graph, but the filleting appeared to work OK.

If you use filletsrf instead of filletedge then you can manually clean up small discontinuities, but the answer it’s a problem with Illustrator is valid, it’s not made for 3d modeling, there is a reason rhino’s standard curves are different.

1 Like

I refit a curve with a few points more than your original and inserted two kinks. Moved the curve seam to one of the kinks and extruded the curve, then capped it.

The curve has minimal radii which can lead to problems with the filleting. You will either have to tweak the curve to be more round or add handles to control the filleting radii.

guitar-body.3dm (597.4 KB)


You could also use SubD tools, which could lead to a more organic looking guitar with way smoother edges…

guitar-body_SubD.3dm (92.4 KB)


Maybe starting by explaining what settings or which procedure you use to export from AI may help to understand what is the source of the described Crvs issues.

Working around a problem that will come again due the used workflow AI > Rhino is probably not the right path.
If a more generic solution exist starting from AI, that would be probably more efficient for the next projects.

I see there are some annoying Knots on the curves for filleting as expected.

Rodolfo Santos

1 Like

Hi Rodolfo-

thanks for the note.

Just saved & imported as AI without any settings being changed.

Side note - been using Illustrator for about 25 years for projects including lots of die-cuts for packaging and displays. First time I’m hearing that a “cutting tool” can’t follow native AI paths accurately.

Thanks Martin,

I’m still really new to Rhino & totally unfamiliar with SubD tools :slight_smile: I guess it’s never too late to look into it.

As far as tweaking a curve, I’ll often be “stuck” with geometry that’s provided to me; I may not be able to change the visible geometry in some cases, so I think your note about adding points and kinks and moving curve seams is something I’ll look into.

1 Like

That’s not the problem. The problem is that Illustrator curves aren’t smooth and precise enough to use as input for freeform 3D modeling operations.


Humm… something may have changed then…

Would you mind sharing the AI file that has been imported in Rhino ( 6 or 7 ?). I would like to investigate.
I use CC, I suppose I will be able to open it natively…

Rodolfo Santos

Looking at your curves, I would say there is no point in using the AI curves as anything more that a general guide, and drawing the for-real curves in Rhino. If you get used to looking at these curves and how much better you can make them using Nurbs curves, you’ll see it will be worth the effort.

I would not do this, if you want a nice thing out the other end - taking your time to get the curves right is the single most crucial part of the process.



This is the absolute best advice one could give.

1 Like

Here you go- (1.2 MB) guitar-body

Thanks Pascal - what makes curves “right” ?

When I say I rebuilt quickly, I didn’t think speed would matter as I simply created curves using Rhino’s tools without much consideration other than making sure they were smooth and that there were no points near each other.


As a first order check of my curves, I always use CurvatureGraph - you should see nice smooth curvature combs, and the continuity on something like this between adjacent curves should be G1 minimum, more like G2 for most. Ideally the curves are single span, or close to it. For more on CurvatureGraph:


Illustrator curves are only “G1 tangent” internally. Every place the radius changes, it changes instantaneously, as seen on the graph where it has a step in it or suddenly switches from one side to the other. That’s problematic for anything that’s supposed to be a single smooth shape, nevermind the little whoopsie at the end where the curvature is spiking off into space(huh, funny, the screenshot doesn’t show the graph going straight out to infinity there, but it is indicating the curve is about to fold over on itself.) The curve on the left is a single Degree 3 NURBS curve, that’s the minimum required for freeform 3D shapes. You see the graph has no steps in it, though it does have sudden changes in angle, when smoothness REALLY counts you use Degree 5+ curves that give you a totally smooth curvature graph.


I would add to what Jim and Better Living said, that right is when it looks right BUT it may take some refinement in your sense of what looks right - this takes some time and experience - The curvature graph is a very valuable tool in helping you see where to make adjustments cause it is not always immediately obvious from the pixels on the screen. So looking right usually ends up meaning nice progressive curves with visually logical transitions and accelerations etc.

My favorite dead simple example is the difference between Rhino’s old Blend command and the newer BlendCrv at its default settings. Both make perfectly good curves that are G2, to the lines in this case, but the graph helps show what is ‘wrong’ with the Blend result - it has two separate ramps up in curvature, and the curvature ‘flattens’ over the largest part of the curve. Depending on scale and material finish, this will show a somewhat abrupt reflection if it is made into a surface. The BlendCrv has one ramp up from either side to a single high curvature in the middle, more like a conic.

To me the latter has more logic, if that is the right word, and is usually what I shoot for. The Blend version is better if you are looking to more closely approximate a round fillet, only have it G2, on the other hand, and I suspect that was the intent when the command was made, eons ago.

BlendCrv on the left this time, Blend on the right.



Wow - great info, everyone. I am learning about some things I’d never known/considered before. I continue to be humbled by the outstanding guidance in this forum.

I think my main question here is how I’ll go about creating curves that “look like they should” based on the shape of an object, but “behave” so that the edge can be filleted properly.

FWIW this file works without issues in form*z:

I think there are two issues, related, but different - your filleting did not work, which led off into a discussion of crappy vs good curves etc, That is a legitimate and even important concern of course in the design of your object but it is not the same as ‘working’ in the sense that the fillet works. I did not ever try your initial file for actually getting the fillets done, but I can do that.

It looks to me like all the fillets are made but they fail to trim the surfaces due to there being two stacked control points here

where it is meant to be smooth. You can see the graph shows a local spike in curvature. AI curves tend to have these, Rhino does not like them. Keep in mind that AI is an illustration tool and not intended for 3d modeling, the requirements overlap but are not the same.

At any rate if that extra point is removed, the fillet does its job.