Rendering a car when you can’t draw
There is a really sneaky way to get around the issue of not knowing how to do a photo-realistic drawing by hand in Photoshop. I don’t have Photoshop. I have GIMP, but even with GIMP and a drawing tablet, it feels unnatural to draw lines. In “How to Render” by Scott Robertson. Scott copies his pencil line-sketch into Photoshop and then uses something called “path” to trace out those lines he drew with pencil. In effect, the lines themselves won’t show in the drawing (because nothing in real life have black lines delineating where a form ends and another begins). With those paths set down, you then have different sections to shade and color. The result is a photo-realistic rendering.
The problem is that in GIMP, the path function is extremely primitive. GIMP, like Photoshop, have multiple layers stack on top of each other to make up a drawing. But unlike in Photoshop, no matter which layer you are on, you can only get that one set of paths, and once you put your pathes down, they are fixed and cannot be tweaked. This difference makes Scott’s technique-the only technique I know-of rendering cars useless.
A little digression: How I got the idea for this project
I was doing this 30-day-draw-anything challenge. I wanted to replace my bad habbit of watching YouTube with something productive. Just 6 days in, I was reeled in by what I love to draw the most: cars. I started by making this little 360 walkaround GIF in GIMP.
(“Welcome to my garage, let me show you around this vintage 1979 Jeep Wrangler Sportback.”)
On the 10th day, I made a rendering of the precursor to the car I was about to seriously render in Fusion 360. The technique has aged really well ever since. Right now, I am looking at it and it still looks sleek.
(My first attenpt at photorealistic car rendering, using the face-painting technique found on Devianart)
Until this time, I still had no idea how I was going to make a 3D model in Fusion 360 that was better than my last attempt. My last attempt of a car 3d model was using the technique found in the video of this free Fusion 360 course. The result looks very dynamic. But I could never find a way to accurately depict the curvature of each panel on the car, like the seam around the window pane. The result looks like something made out of Play Doe, lacking strucure and organization. Well, what can you say? ;P I didn’t come into this project with any blueprints.
(First first ever original 3d car model)
I don’t like the last technique. I want strucutre but also originality. I need improvement. Then, I remembered long hidden in one of my YouTube playlists is a tutorial by Khelifaoui Ammar on 3D modelling using a different method. This time, instead of tweaking a huge surface here and there, which makes it really hard to get an accurate overall structure because a tweak in one place can affect the geography around it, I will be tweaking small distinct surfaces. What is better, I can choose where these surfaces are. To add to the motivation, I will use my own design.
(my intial sketches and breakdown of the surface)
End of digression
So with the Scott idea dead, I gotta find a way to render a car realistically. I didn’t go to art schoool. I could do somewhat a perspective drawing, but not quite. Then, this idea of using Fusion 360 just hit me. I’ve been using it for some time and it got a super neat Rendering feature where not only the surface looks real with reflections and all that, but the whole car would be in perspective. So it means if I build this, I can use the resulting model to learn from and sketch from. It’s like I can finally have a way to completely visualize what is in my head. My hand worked well in term of sketching out the car I am thinking, but I guess my skills aren’t yet mature enough and I don’t have the tool (Photoshop) to do as well a rendering as Fusion 360 can.
I pulled out Ammar’s tutorial from the dust-covered back log of 3D modeling tutorials I have mindless saved over the years. Following it, I first drew some front, side and top orthographic reference images (meaning no perspective, foreshortening, etc. involved). As I 3d modeled some more, conflicts arose between these images. For instance, when I stretched the length of the car to be consistent with the length of the car in the side ref image, the body appeared longer when viewed thru the top ref image. Along the way, I 3D modeled and tweaked my ref images so that they were consistent in the 3d space.
(The front, side and top orthographic reference images I was using by the time I completed the model)
It took me 21 days to go from nothing to this:
The result wasn’t perfect. There was no side-view mirrors. If you compare it to the photo-realistic rendering I showed before, the model lacks some body sculpting, the side and front skid plates. My plan of putting a bloody glorious air diffuser got only as far as a white plastic cover on the rear.
But the bright side is that this is DONE. DONE. OK? I mean, DONE. That was 100x better than something that would otherwise have just existed as an idea in my head. The result that now I can see feels so motivating. This won’t be my best work, but it will sure be a stepping stone to it.
What could’ve been done better
First before I say anything, I want to say I love this car. The tutorial I followed uses a BMW. But I made the project even more exciting by making a car I designed. But coming into the 3D modelling process, I had no reference images pre-drawn. That, I wish I could have spent more time on. What I should have done was to draw out all the orthographic references images first (finalized but still open to changes along the way), overlay them on top of each other and make sure all the features, seen from different sides, align. For example, the windows on the two sides should be of the same length when seen from the top and from the side.
Secondly, I wished I kept it simpler and more to the point (as in serving some sorta function besides just looking good). Keep it simple and stupid, basically. The rear air diffuser is so complex to model and look out of place on a SUV that aren’t particularly sporty as suggested by its styling. I think keeping the look simple will result in more satisfaction when you end up modeling all the features you designed.
Lastly, I wish next time to try a little bit with designing to solve a particular problem. Yeah, start with something out-of-this-world but think a little more about how the look will support a function in some way. Otherwise, no one will care about it, not even the guys on Twitter….
….Where was I….Anyways, I hope you learn something from this article. If not, tell me why in the comment and I would love to know.
