Spent a day and a half looking at the new release of Eyeon Fusion with Matt Leonard of SphereVFX. Nigel Richards of the distributor FocusFX arranged some tuition and Framestore donated their training room. Matt has been looking at Fusion betas for a while and has good experience of the Nuke package, so he was able to give a balanced answer to most questions that came up. He doesn’t know the scripting side of things that well, but showed everything else I wanted to see. The new 3D and GPU render effects are fantastic with multi layer rendering, normal map handling and the stereo facilities look interesting.
The new version of Fusion gives a lot more information in many areas, the meta data from images, information about the tool settings etc Doing a click/drag from the colour picker button shows not just the colour info while dragging across the screen but the alpha channel values. If you hold the Control key down while connecting the output from one tool to another, all the possible inputs to the tool are shown. It all makes the software easier to use.
A major part of this upgrade is to the 3D tools and they have many more facilities than can be covered in a short blog. The system is able to read in fbx files from any of a number of 3D packages. It can texture, light and reflection map these models and even write them out again for processing in other compositors. There are a number of shading models; the usual Phong and Blinn with the addition of Cook-Torrance and Ward shaders give a wide range of finishes. With the ability of being able to mix shaders, the system is able to give complex shaders from simple plastics and metal through to car paint, skin and wetness. The 3D also applies to the various types of stereo tool available. Stereo can be produced from a single camera by adjusting the eye distance and focal length, or more complex situations can be created by using two cameras and combining their viewpoints. It looks quite comprehensive, but I’d need to get a real project in to see where the practical benefits are. The 3D upgrade is really impressive and Matt showed us some good examples but it needed more time or again a real job to show its power. Features of note include the fog, openGL rendering, fbx exports, texture (uvw) transforms, normal mapping 3D tracking etc etc. Really impressive.
Two phrases have been brought in with this release, “Domain of Definition” and “Region of Interest”. The Region of Interest is a way of just processing and displaying the part of the final image that is important. Only this area of the screen is updated through the whole composition, so it makes the render much quicker. It’s as quick to do as clicking and dragging a rectangle around the area and so should be really useful. The Domain of Definition on the other hand, is a cropping of the area being processed by a pipe, so that no rendering is done outside of that area at all. Material outside the area is discarded. This can be done at multiple places through the pipe and again, will improve the rendering times proportionally to the area involved. Effects can be limited to just a small part of the image. Where these are quite subtle to see, a tint can be put onto the video to show the area without the effect. Useful.
Splines have changed a little in the way that they work. Matt talked about the use of beta-Spline rather than b-Splines, giving more control. I seem to remember that changing points and tangents with bezier spline functions used to affect the curve as a whole, whereas beta spline functions only affect the curves connected to that control point. They seem to work well; they have a few extra keys to create the double splines for smooth masks, move groups of points together etc. They also have a useful mode to affect all keyframes. An extra smoothing point can be added to all key frames at once, rather than having to go back and repeat the action to all the keyframes already made.
I haven’t really looked at Generation, but I’ve seen that the two packages can be used in conjunction to give more editing and versioning features. If you don’t want to bring in an Avid, it can perform most of the basic editing and colour processing. It also deals with versioning of a composition. If a comp is double clicked in Generation, it will open for editing in Fusion and will save as a new version, but it doesn’t destroy the old version. The user could revert if they prefer, to one of several prior versions of the comp. Fusion audio hasn’t changed much. It is still limited to just a simple scratch track. It could do with some extra audio processing facilities, even just some simple sound mixing would be useful. Chroma keying is much the same. Keylight is probably worth buying if you need that extra quality.
All in all, the two day SWAT course was really useful. Matt’s experience with other packages gave an insight into where Fusion was ahead of the competition and there were a few hints on where it might catch up in other areas over the next release or two. If you need to get up to speed in this version, I can throroughly recommend his DVD which has several hours of tutorials on the latest changes. Even after being on the course, I found the videos showed yet more of the detail of the new features and I’m sure I’ll replay them several times over. Another really good version 6 update descripton is on Matt’s entry on fxguide.