Things to think about during a VFX production Adam Avitabile, VFX Supervisor, Look Effects
Unfortunately I missed the beginning of Adam’s talk which was about their work on Step Up 3D. When I came in he was talking about his experience as a VFX supervisor and giving a few tips he found useful…
- Stay near Video Village
You never know when the director will need to talk to you so you have to make yourself available. You should show your presence.
- Get to know the crew
If you have a good relationship with them you will be able to count on their help… and this might well save your life!
- Never tell the director “no”
Instead you should explain to him/her, what problems whatever he/she wants to do will entail and try to find a compromise.
- Get to know the editors
They are the first to see your work and give criticism. If you have a good relationship with them, they will trust you more and try to help you.
Stereoscopy in Film Production: Theory and Practice Parag Havaldar, Software R&D Supervisor, Sony Picture Imageworks
I know nothing about S3D so I was really looking forward to understanding the process a bit more. This conference was very interesting and gave a good overview of what this new technique entails.
- For fully CG films, you just have to render everything twice.
- For live action, you can shoot in S3D but you will also need to make a lot of corrections in post (alignment, color, lens distortion, vertical disparities etc).
- You can also convert to S3D in post. It helps to know beforehand that you will be converting to S3D, as you will be able to gather mode information on set that will help with the process.
- There are also legacy projects, which are the conversion of older movies (Titanic, Gone with the Wind) This is mainly done to generate new revenue with a new theater release.
- Finally there are hybrid projects in which the post production techniques are used to correct a badly shot S3D scene.
A few technical points:
- When you increase the distance between the two eyes, you increase the overall depth of the whole scene.
- You have to be very careful when setting the point of convergence. This is the point where both eyes see the same thing. It is also where the viewer will focus.
- To create S3D in post you have to first isolate the different elements. You can then create depth and render the other eye. Finally comes the clean up process.
- You can also create the second eye by using computer vision techniques. But this does not work for all shots. You need to have very high level of detail textures. Indeed, the algorithms look for correspondences in the images so if everything looks the same, they will not work properly.
Renderman workshop Dylan Sisson, Renderman Technical Artist, Pixar Animation Studios
I didn’t manage to attend the Renderman presentation in Annecy this year so I’m happy to be able to here.
Dylan Sisson started off with an overall presentation of Renderman and rendering at Pixar. It is specialised for feature film rendering so depending on your scene it might not be the fastest renderer. At Pixar they usually render everything in camera (no compositing). They only split a shot to speed up render times (not everything has to be rerendered for each frame). Here is a graph of typical render times per shot and their evolution through time.
Dylan Sisson then talked about one particular problem they encountered on the production of Toy Story 3: trash. In the incinerator scene, some frames could take up to 40 hours to render! They are very complex shots with tons of geometry. The strategy was to divide and conquer: i.e. use different methods for the foreground and the background. For the foreground they used more advanced techniques to get interesting effects like point based color bleeding. For the background they used one general scalable shader. They could control the density and the type of trash directly in the shader. This is a very cool feature and must have really facilitated the artist’s work!