Hello everone and welcome to the exciting tutorial of integration 3D elements into live action footage. In this tutorial I will go through the steps nessesary for creating an image on top and important notes that need to be taken into consideration on all stages of the process. I've done 3D integration several times now and results are getting better and better so I hope you find this tutorial usefull. The model that was integrated wasn't modelled by me and was taken from models collection. This tutorial is the process that is documenting the assessment for the paper of visual effects and has an educational purpose only. There is no commercial interest what so ever.
Here is the final render of the project so if you follow along you should be able to get something like this digital outcome of convincing integration and tracking pocesses.
Anyways, apart from other things I used several tools in order to be able to achieve this result. I used photo camera Canon 5D Mark II for taking the footage, then tracked it with Boujou and then integrated 3D object using Maya. In order to achieve high level of realism I used the image uwrapped from the reflectiive ball using progamm called Photomatix, done some setup of the composition in Nuke and final compositing was done in After FX. The entire process was done on Mac OS X. I hope you enjoy this tutorial and if you have any questions or examples of your achievements, please flick me an email on email@example.com and I''ll be happy provide you a feedback on that. Cheers. Have fun!
First of all let me explain why am I chosed this exact model. Well for me its aone of the most georgeous car that men had ever built and designed. Secondly the shapes, textures and forms of this model basically attract more attention than the environment, thats why I though that it will be a cool object to place because the audiences will watch the car rather than finding problems of integration process.
The idea basically came from the ability to place CG image into live action that it is about placing something artificial into the place where the CG object couldn't exist normally. So I decided to place a car into the environment where it cant drive to or located in any way.
First of all I made 4 hand made markers for better tracking.Also measured the distance between them for better scale and realism.
I found the spot to place a car that is close to my house and has spectacular scenery and also the car can't drive there.As you can see on the image below there are 4 markers placed on the ground to represent the car shape and also provide a better tracking results for integration.
As you can see the ground is not flat that is exactly why I needed those 4 markers and you will also see later that the tracking went way better with using manual placed markers.
Before the live action plate was tracked I went into Nuke and exported an image sequence for importing that into boujou.
This image represents the way BOUJOU placed tracking markers on top of live action plate and tracked on the frame by frame basis.
This is the camera solve in Boujou and as you can see what camera is doing in the scene. Its pretty much moving around the round spot ... or around the future placed car
I used high sensitivity for tracking the footage.
Then I've imported the tracked footage into maya and set up the camera settings for rendering.
Then I've placed CG car into the place and made playblast tests to make sure the CG fits the space of live action and look as it was actually there.
Then I went and started my HDRI pipeline using reflective ball and two stable tripods. I took about 20 images with different exposure values for creating the full range of light.
I used the Photoshop on mac side for combining HDRI information alltogether into one image.
Then I cropped the image and saved as HDRI.hdr file 32 bit and cropped the ball for later unwrapping in the sodtware for mac side called Photomatix.
Next step was to unwap the ball into one panoramic image.
Then I mapped the image on the sphere in maya and used it for defining the lighting information.
Its very important to make sure that the image wraps around the placed 3D object as the object was placed there so the reflection represents the exact relationship between the object and the environment.
Usually you can use the cylindrical UV mapping in order to rotate the image and rise up the ground level.
So this is how the final scene look like after integration and assigning HDRI image to it as well.
Here there are some rendering results that I came up with so far tweaking render settings and lights as well. Used mental ray and maya to render final images.
As you can mention from the image below the reflective ball was placed exactly at the area where the CG car will be placed on later so it has perfect relationships with the environment.
Then I've made the final assignment to the setup of shaders and directional light setup, on the image below the final render with mental ray render and mental ray shaders. Highly reflective shader of the CG car has the mapped HDRI reflected on the surface that enhance the realism of the placed object into the scene. As you can see there is no shadow from the car yet which was quite problematic task to accomplish later on.
For the sake of adding realism to the scene there were tires of the car deformed as if the car had weigh so the air pressure in the wheel decompress the tire as in real life. The front wheels were rotated 35 degrees as well as the steering wheel inside the interior of the vehicle.
Here is the steering wheel rotated accordingly as it usually is in real life.
The next step was to move into setting up the render passes for the scene. On the image below you can see the alpha channel of the color pass that will help to pull off the matte of the image for better control and convincing compositing later in the digital compositing application.
And this is the color pass again of the same frame but now there are all three RG and B channels turned on.
Down below there is a ambient occlusion and shadow passes.
As you can see the area with the render passes set up and consist of 5 different layers.
Very good advice for working with a massive amount of the geometry data in the scene is to keep the scene view at the bounding box option in order to have a better saving time and view port performance as well.
After long rendering passes eventually I jumped into the After Effects and Nuke in order to sep up layers and channels plugged into them in order to compose them properly. First I've managed the compositing in AE in order to back myself up in case the Nuke will not be easy to use, but then I figure out both compositing applications.
On the images below you can see the setup I used both in AE and Nuke of different render passes. Also here is the render of the explosion which I will going to use in the further developments of the actual project.
Fo the sake of using this video in my show reel i put the software I used in order to create such composition.
So those are the final renderings of the footage and compressed CG object together. I've added some Fast blur for some areas of the image and also a lens flare again for the sake of realism. Further on this project could be enhanced by adding more layers for it such as particle effects or liquid simulation or any type of destruction and explosion as well as adding one more shot in order to move narrative forward. In this case the callenge will be to make background plate and CG object look similar to the previous shot but in the same time add one more beat to the story.
Well to be honest this is the end of the tutorial and I hope you enjoyed going through it and following along as well. I hope you will get back to me and my website to have a feedback ot just to have fun. The truth it that I'm going to continue the development of the narrative for this project and in the next shot apparently there should be a moving CG object which migh be be a driving car or explosion or transformer or whatever I will decide to make this more effective and beliveable. Hope you've enjoyed the show and wait for the continuation of it. Cheers. To be continued. :)