Skip to content

Work in Progress 3 – Link Constraints

I have carried out an in depth analysis of the reference footage, I would estimate I have watched it at least a thousand times. The work of other amateurs I found in my research is interesting and helpful to see, but nothing compares to watching the video produced by the film studio.

I have picked up a lot of tricks used to make the job of animating it as simple as possible. For starters, the entire shot is framed from the side, essentially a 2D shot; only the side of the car can be seen, which means essentially animating a silhouette. The camera zooms in so that for the first second (the part where the most action happens), only the front end of the car is in shot. The artists heavily use people running infront of the camera to mask the unrealistic movement of car panels. However, despite noticing these tricks, I have opted not to use them, simply because I believe I have the time and skill required to do it without. This will give me a better final project and improve the quality of my showreel.

I have also made some deductions myself: Not all parts of the car need to be used. Parts can be scaled and skewed as needed. Some parts of the final transformer, which appear to come from the car (same paint job as car bodywork), are custom created.

Below are a miscellaneous collection of work in progress screenshots, following that is a write up of my progress, including rendered videos.


This has been a long and tedious stage of the project. After the artistic decisions of how to break up each panel of the car, and where each part should go on the transformer, actually going through with it and making my vision a reality is dull and repetitive. The placement of object pivots is an important stage, which could make or break the animation of the part; a well placed pivot means fewer keyframes and a smoother effect.

The placement of the car parts onto the transformer was done with link constraints and auto keying, so this process alone yields some rudimentary animation. Because of this, the placement of the link key frame is extremely important as it can save or create the need for extra animation work.

Each part is motionless until the animation slider reaches the link key, at which point it is attached to the bone, and as it keeps its initial displacement, needs to be auto keyed into position, that key frame is then copied and pasted to just after the link key. This means that the part will seem to be part of the structure and move with the bones from this frame onwards.

For each panel, I spent a good amount of time thinking about the earliest it can become attached to the skeleton. Any mistakes, of course, can be corrected, but the movement of the link key frame would dislodge the panel from its carefully aligned position and that work would have to be done again.

When it comes to the actual animation, for the most part I have just placed key frames and let the algorithms behind the scenes do the rest. My requirements of the curve editor have been relatively limited. However, when the movement of a particular part hasn’t been as smooth as I would have liked, I have used it without hesitation.

I have compiled all my renders so far into videos to save space and make for a more fluid viewing experience. My modeling and lighting progression can be seen below (change quality settings to 720p):

You will notice that the quality of the lighting setup improves drastically in the last few frames of the above video. This is because it would have taken me too long to render in 720p every time I made a small change. I used a very low resolution render quality for this stage of development, which can be seen in the following video:

In this lighting stage, the challenge was to make sure the sequence looked good, both as a car and as a transformer. As the latter is about 4-5 times taller, this required the construction of a large scale set and the lights placed quite far away. It was quite a fiddly process but I am pleased with the results.

Here is a medium quality render of the project once all the car parts have been linked with varying degrees of animation; some parts are animated well, others not at all with intersecting geometry etc. (change quality settings to 720p):

For all aspects of the pipeline of this project, I have experimented with a number of ways to achieve each thing before developing it fully. This makes it easier for me to avoid making bad decisions based on an attachment to something I have invested a lot of time in. I feel this is a very important discipline when working solo on such a large project.

Doing my testing early also allows me plenty of musing time to come up with new ideas, it gives me an idea of the time I will need to devote to each stage of the project; it means I will not be left to rush any of the later stages of development due to deadlines.

I made the decision to change select and link to link constraints. To date this is the most important decision I have made, and at this stage I feel that, without it, the project would have failed before it began.

I made the decision to change the lighting from an HDRI map to a studio lighting set up. A perfectly clean car felt out of place in an outside setting.

I made the decision to use a more exposed camera angle. All of the car is in shot at all times, and I cant hide panels moving behind people running infront of the camera.

This project is my artistic vision and I make all of the calls. All of these decisions were mine. However, I was led to some of them based on my review of feedback from my peers. As they fall comfortably within the target audience demographic, their opinions are valuable.

The extensive testing I have done before committing to any particular technique has been absolutely crucial to the success of this project so far. I had never really before understood how to critically self-evaluate my work. In this project, I am always taking a step back to look at my work and ask: ‘Is this the best way to do this?’ A lot of the time, I don’t know the answer to that question, so I try a different method to find out.

Work in Progress 2 – Troubleshooting

Having to lower resolution from 2.7mil – approx. 250k: Most of the car parts had 2 iterations of turbo smooth on them, but the stack was not collapsed, this meant I could delete the turbo smooth modifier and manipulate the low poly mesh. However, some parts were collapsed to do some poly insetting/bevelling, ie the badge on the front bumper. I had to import parts from earlier saved iterations and merge the two parts. This served as a good reminder of importance of saving in a new file as you work.

It is extremely important that panels are positioned perfectly when they are linked to the transformer, any movement of panels will cause problems when they don’t align in original frames. It is very easy to make mistakes; I have found myself constantly turn off edged faces and checking around car. Subtle things such as wheel height that might be noticed later will be difficult to change later when lots of animation tied to it. I adjusted it so that they were perfect before linking.


My method of linking was poor; it left panels not perfectly aligned even if the pivots were unchanged, and the motion of the skeleton affected the interpolation between key frames. Initial animation testing methods were all on one side, so when it went wrong I just deleted the geometry and mirrored over new geometry from the other side. Worst case scenario would have to import in parts from old saves, but havent had to do that yet.


I started using link constraints – this is a far superior method. I have redone animation on all parts. It is a much more secure way of animating also, avoids any panel movement providing I don’t forget to turn auto key on and off. My method was dissecting the car and slowly linking pieces one by one, always checking that the original position had not been disturbed. Placing them on the transformer in their final locations, copying set position keys to new locations so that they can be selected without disturbing the link key.

Half of the animation in between will be unlinked, real world animation; so that I can make it realistic, without being hindered by the moving skeleton – which makes it hard to animate as it affects the curve in between two key frames.

Another problem I encountered was panels that had been broken and then smoothed were breaking up smooth shapes, my solution was to delete turbosmooth on both, attach and weld. apply turbosmooth and collapse, then separate into different objects. some distorting still but better than before.failing to spot this would mean spending hours animating these two panels only to notice that the only way to fix them would involve deleting all that animation.

It is crucial not only to solve problems but to anticipate them too. This might seem like overkill but it is this attention to detail that, I believe, will make this an exceptional showreel piece.

Comparing renders – how something looks in the viewport is not always how it will look when rendered, for this reason I frequently render and check that everything is in order. Every time I break a piece of geometry and make new seams which I would like to be hidden, I render to make sure the shaders are behaving as they are in the viewport. This is the attention to detail that I put into my work, to ensure it is the best it can be – of the three changes made here you probably would only spot the crack in the panel being repaired:

test10 test11

The other two changes were the grounding of the rear wheel and the improving of a hidden seam directly above the front wheel.

Found a problem where I was halfway through linking car pieces to model (necessary to complete the modeling) also testing animation, both of pieces and of the biped. Noticed that autokeying of biped parts to improve quality of standing up transformation would adversely affect any piece of the car linked to that bone. ie anything with link constraints would not stay in position when trying to keyframe bone positions within the confines of the keyframe information on the linked pieces. I experimented extending workspace past my safe t pose frame and animating bones there. linked pieces worked as intended as they have no future keyframe information at this point. option 1 looks like getting standing up animation perfect and then removing all the dislodged parts. discovered that for things like spine pelvis etc that requires rotation, i could make a rotation keyframe (eg -10 degrees x) in my safe area past my t pose frame, and then move that ontop of or inbetween existing key frames without affecting car geometry. can also delete key frames without it affecting the link constraint pieces.

This problem only affects first 4 seconds as thats how far I have keyframed. For later sections of animation, I will finish all primary animation first then secondary animation. The head was unaffected as no link constraint parts. Elbow look at constraint also unaffected fortunately. Not many parts put on legs at this point so not too much hassle there.

The solution I employed was to copy and paste key frames to the safe area and building new animation there then drop it back in. Model is all over the place so a little confusing but a solution nonetheless:

Troubleshooting03 Troubleshooting04 Troubleshooting05

Another excellent demonstration of the importance of testing ahead of schedule. I still havent finished modeling yet but am fully aware of what is going on in all of the next stages in the pipeline. I have been doing early materials tests with placeholders and thinking of how many different ones needed. This means I am getting a lot of time to look at it in test renders etc and make artistic decisions. come up with different ideas of easiest way to do it.

Current idea – uving some portions, have a few different hi quality general weathered metal textures in photoshop, make from them all the various maps required. use same texture repeatedly on different parts, no 1 big unwrapped texture map. just all the individual parts unwrapped on top of each other; each part using the whole texture space.

When first rendering to see what it looks like, I analyse and break down all elements to see which direction to go with them. Needs to be a camera angle from which it looks best although I would like to make it acceptable from all angles.

I have produced a few draft quality renders of my progress so far from a number of camera angles.

Evaluation Plan

As my gannt chart shows, I aim to complete the final product significantly earlier than the deadline, so that I have time to react to feedback and implement changes.

The product will then be shown to both peers and tutors, from this I hope to gain useful critique from both a technical and aesthetic point of view.

Although I am already ahead of schedule, if the product is not finished by the demo point, I will show what I have rather than push back the deadline. As it is important to get feedback at a point where there is still time to make changes. I will then schedule a further screening of the final product when it is completed.

I have been constantly testing the product as I progress, developing multiple methods to do things when I am not sure, then taking a step back and critically evaluating my work. I have been testing all aspects of the pipeline before I get to them. I will be continuing this practice throughout the project, posting my work in progress online and seeking feedback from peers, tutors and a variety of appropriate forums.

I have also been contacting people I know to see if anyone has any links in the industry. A friend of mine has a relative who worked on the transformers films and has agreed to put me in touch. This would provide great help with my evaluation.

I will also be creating a survey using a website such as I will be seeking independent advice to assist me in the construction of unbiased questions that will yield useful results.

The final product will be uploaded to Youtube and Vimeo. I will post links to social networking sites such as Facebook and Twitter, as well as the forums.

My progress from start to finish will be fully catalogued online in my blog, which will be open to the public, and I will be drawing attention to this on the aforementioned mediums very early in the development process. I will also be creating an official website which will house all of my completed projects.

I will be releasing ‘teaser’ clips and stills at various milestones along the way and advertising these appropriately.

The product is being created as a show reel centerpiece as a tool to gain employment; self promotion is just as important as product promotion. I will be promoting myself alongside the product, developing coherent branding (including a logo) early so that I have a consistent and recognisable image when I start to release interest-generating content. All work I have online will be updated to include branding and all new work will include it.

Time Plan

I have completed the rig and made IK controllers for it. I have completed a large portion of the modeling. I have retopologised the car and started to break up and solidify the panels of the car. I have also done a lot of initial lighting, texturing and animation tests. This is my time plan for the remaining weeks.

Gannt chart

Work in Progress 1 – Initial Testing

The following pictures include reference and inspirational photos, work in progress screen shots and renders.

Moodboard Rig Head Foot Materials

Below are some video test renders. These are by no means a representation of final animation quality, they are strictly tests to gauge how much work I have ahead of me. The first stage of the animation is to animate the camera. I have watched breakdowns from the transformers films, which show that they animate only what is in shot.

Storyboards/Job Spec

Skills Required: 3D modeling, retopology, rigging, lighting, texturing, animation.

Job spec: Model, rig and texture a transformer to a target of <200k polys, retopologise car from 2.7m polys to ~200k polys, rig car to function as a car and as transformer, animate sequence, lighting, rendering and concept art.
As I am trying to reproduce an existing video, storyboards aren’t particularly important to my progress, as I will be using the video for my reference. However, my brief dictates that I should produce some so here they are!


My research consisted of just looking at what has gone before and thinking about how I might do it myself:

Something I noticed when looking at these videos is that most seem to be weak in one area; those with good modeling and texturing have poor animation, those with good animation and rigging have low quality modeling and texturing. It is important to me that my final product is strong in all areas; it would be a shame to waste a good model with wooden animation.

It is also worth noting that ‘animation’ covers two different tasks. The first being conventional human animation – standing from prone, this should be as realistic as possible. The second being the animation of all the parts and panels – this offers much more creative freedom, however, I think an aesthetic result comes from maintaining some degree of flow to the transformation so that the viewers eye is naturally drawn around the piece.