Blog

Dragonflyfeatured 620x183

The Making of : Mechanical Dragonfly

Introduction

This ‘making of’ aims to show you the general process involved in taking a concept drawing through to a final rendered still. In this instance, the subject matter is insects. I have had a lifelong fascination, and Dragonflies are right at the top of my list. They are very powerful insects, with unparalleled hunting ability (they are one of few if not the only insects that can fly backward as well as forwards, sideways, and hover).

The idea was to recreate this celebration of natural engineering in a manmade form. Enough of it’s little intricacies can be seen without making too unwieldy and heavy to work with. Insects all follow the same anatomical rule. The body is divided in to three distinct sections (head, thorax, and abdomen). They have three pairs of legs, and one pair of antennae. This all suits the CG creation of the creature very well indeed. They also have a pair of compound eyes, and ocelli (secondary eyes). Organics not being my strong point, I went for a bright daylight tone to the final render, with just sky in the background.

Concepts

Rough ideas to designing the model

Step 1 : Initial Designs

I was given a selection of rough design guides for the Dragonfly, drawn up by Steve Tappin. I had free reign to choose the design(s) I liked best, and decide how to put it all together. My main desire was to have a heavy powerful look, like the real thing.

 

 

Step 2 : Model Rough

Having settled on the fourth design shown, I needed to do a quick mock up to see how the body design would look in 3D. This is a pretty quick step, as no real concern is shown with how it’s modelled. Literally it’s just bringing shapes together to replicate the form.

 

 

Step 3 : Modelling over the Rough

Having determined that the proportions and shapes essentially work ok, the initial modelling phase is a model over, covering the rough with the starts of the final geometry. This model over phase isn’t long, but is a useful guide, avoiding some guesswork early on.

 

 

Modelling

Building up the Detail

Step 4 : The Head

As insects are modular in appearance, modelling up the body was done in sections. The head as with the rest of the model was poly modelled to be subdivided. The detail was gradually increased, with an internal cage structure being added right at the end to be seen through the body material.

 

 

Step 5 : The Jaws

The jaws were the most complex part of the model, being fully articulated with dual piston hydraulics as well as a rotational hub. The idea was to make them cool for possible animation. The mechanism is modelled to allow it to fold flat under the head when at rest.

 

 

Step 6 : The Thorax

The thorax started out simply modelled over the model rough, and using the stencil and edge bevel tools, details like torx bolts were added, as well as the grooves in the sides. To add more visual interest, intake vents were added to imply there was an engine requiring oxygen.

 

 

Step 7 : The Abdomen

The abdomen sections were detailed up from simple tubes, scaled to give them shape. Initially I added groves down the side to match the thorax. With internals added to other parts of the Dragonfly, I changed the design to a split shell with internals and an enclosing frame.

 

 

Step 8 : The Legs

The legs were pretty simple, poly modelled again like the rest of the Dragonfly. I added an internal frame here as well. The wings were again poly modelled with a single poly modelled membrane in a metal frame. If I had a knowledge if zBrush, this could likely have been used in various places.

 

 

Lighting and rendering

Putting the final touches in place

Step 9 : Composition Testing

Early on in the overall process, I had made a test render which as soon as it finished rendering, I liked the dynamic of the pose. It mimicked the way Dragonflies tend to rest on known regular perches. I would need to add said perch, but the basis for my render was definitely there. This was to be pretty close to the final image composition.

 

 

Step 10 : The Backdrop

The backdrop was intentionally simple. Rather than risk organic modelling, I created a simple blue and yellow skewed linear gradient, and a second layer with a same coloured circular gradient. I screened this over the base gradient to produce a nice simple angled hazy sun and sky.

 

 

Step 11 : Shading Test

Shading was pretty torturous for me, as I really needed nodal shaders to get the translucent effect I was after. Simpleskin seemed the most successful in general, and this was actually one of the better results. Nodes are undoubtedly the way to go, but not easy to grasp after 10 years of layers!

 

 

Step 12 : Ambient Occlusion

Ambient Occlusion is a great way of adding depth to a render. Smaller details are picked out nicely by the arbitrary shadowing that is created. More use could be made by rendering passes, but as a modeller I’ve never really delved that deep in to the use of passes.

 

 

Step 13 : Putting it all Together

The AO pass is overlaid on the colour layer. The alpha is used to clip the colour and AO layers to allow the background to show. I then flatten, created a blurred layer apllied as a 18% colour dodge. Then a noise layer added, followed with a vignette, and a colour overlay tint. It all builds up to degrading the CGI look, and fooling your eye in to thinking it’s real.

 

 

The Finished Article

With a final rendering time of approximately 1.5 hours rendering at 1080P, it’s a pretty good result. The specs of the machine used were a quad core Intel 2.4ghz, with 4gb of Ram.

Additional Thoughts

Shading & Nodes

An odd design revision aside, the modelling was pretty straight forward. With my limited knowledge of nodes in Lightwave, they are clearly very powerful, but they also can be a real rendering burden for some of the cooler effects. I used SSS nodes that gave a great blurred refractive look, but the render hit was impractical. Several of the nodes used for Skin produce very similar results but with wildly different render times. My lack of experience in knowing how to use them fully was a big hindrance. Learning nodes is definitely a worthwhile investment. It WILL pay dividends!

Leave a Reply

You must be logged in to post a comment.

If you have requirements for modelling, contact us today!