The second demo of the afternoon at the 2003 Discreet users group was a technology demonstration of new stuff being prepped for the next release of flame and inferno. Not everything that was shown at the show will make it into the software — potentially the features will never make it in — but it was an impressive showing nonetheless. The software engineers, who just wrapped up the 5/8 release and maintenance releases, pulled out all the stops to show s
The full report:
Clip History
Tops on the list for most users (and our fxguide survey) was the showing of clip history features. Product specialist Martin Helie, known in some circles as the mayor of discreet, started out the demo by taking a clip through various desktop modules….keyer, color correct, logic ops, etc. And then -on the desktop- was able to pull up a list view of the operations that the clip had been sent through. In theory, one could go to a single item on the list, make modifications, then redo all the later steps automagically. But even cooler, he took the final result clip into batch and then “exploded” the clip into a batch tree which reflected the history. Finally there is hope that our result clips will include info about all the steps that were used to create it. Say goodbye to super-clever-long-naming-conventions.
Batch Improvements
Using the history tree as an example, the next batch concept shown was an improved group node. A user could take a selection of nodes in the schematic, group it into a single node, name it and then drag the node down to a new special fourth row of nodes in batch that contains custom nodes. They look the same as as regular batch nodes but contain the various sub nodes….basically a way to roll your own nodes for repetitive operations.
Helie also showed the concept of batch node aliasing. If you have one node, you could make an alias (linked copy) of the node and when you adjust one node it adjusts all of the other aliases. Nodes would be distinguished in a visual way so that users knew which nodes were aliases and which were linked. Think of it as automated expressions in the new batch, linking one node to another but in a cleaner way.
Uber Keyer/One-Touch Keyer/Keyer of the Year
Another item shown at the users group was a new discreet keyer, which was quite impressive. The long-haired cat on bluescreen made its annual appearance at the users group, but the key was much easier than any keyer demo in the past — even considering that such demonstrations always seem to key footage magically. Sampling the background color in several areas quickly pulled a basic key — seemingly much better than other keyers. Sampling pixels in foreground areas and background areas intelligently turned the matte pixels white and black.
Spill suppression and color correction was automatically applied to adjust the foreground image (based both on the backing color and the background image that the foreground image was being keyed over), and the amount of change was adjustable. Grain tools for keying were built-in. Helie showed applying a degrain-type tool to the image, which actually brought out detail in the image that was previously difficult to see. In the example, it actually made whiskers keyable that were previously not even visible. Also, there were considerable edge adjustment controls in order to more seemlessly blend the foreground image with the background. Matte curves on steroids, to say the least. All in all, we can’t wait to get our hands on this new keyer — whatever it ends up being called.
Spline-based Warper
Yes, after years of hoping a new warper wasshown. With seemingly no one making a warper any more, this would be a nice addition. I have to say I’m tired of using extended bicubics to do warping… What was cool about this demo (in addition to the obvious fact that a new warper was shown) was the way you could draw a spline. Helie showed that you could simply draw a freehand shape with the pen and have the software interpret where to put points and tangent handles. After drawing, you could adjust the number of points on the curve, making it more detailed or less detailed. Hmmmm…the splines looked strangely like GMasks — could this functionality be coming to gmasks in the future as well……????
Burn
Really not much to say here, other than the fact that there is a “Burn” button in the UI which will send a batch render off to a networked burn software renderer. It no longer uses hardware as was previously shown at past users groups. The software saves the source clips on a reel in a new auto-created _Burn library and the resulting render is also saved on a reel in the library. A PC application monitors burn rendering and status. Speed alss seemed to be quite impressive, though in private conversations with discreet, benchmarks are still being completed.
Photoshop Import
Photoshop import was also shown at the users group meeting. A Photoshop mockup of JFP’s business card was shown in Photoshop and saved as a .psd file. In the flame software, there was the ability to import all layers, selected layers, or a flattened image. The demonstration file had a solid layer with transparency as well as a text layer — both seemed to work fine. Helie alluded to the fact that future considerations might include importing the Photoshop file in action, so that all layers, blending modes, and positioning information would be available and ready for adjustment. All additionally made possible by the multi-resolution stuff that was just added in the recent releases. Yeah….that would be cool.
Mental Ray Rendering
The last bit of the demo was also incredibly impressive — the showing of Mental Ray Rendering of action images. The example shown used the imported photoshop file and various lights to show cast shadows, translucency, and more. The render of the image was actually quite speedy considering the calculations being done. The implementation considerations would be quite interesting (aka — how do you handle all of the shader values in the flame UI) and its not meant to replace a true 3D system, but the usefulness of this shouldn’t be underestimated. Bring it on…
Little but Cool
Last but not least some cool stuff that was shown….not biggies, but…
– Logic Ops has new modes (like Divide — woo! woo!), a matte input, and a mix/transparency value. Mix in batch!
– Parenting across multi-views. If you have one side of the schematic in one view and another side in the other, you can drag a parent line across the dividing line between views. Now that would be a really cool way to link across different action nodes in batch (though I suppose you can’t see two different schematics in batch so it wouldn’t be possible oh well.)
– Handles on timewarps from EDL assembles. Yes, coming soon to an inferno or flame near you — define the frames of handles you want on a timewarp in an EDL. Based upon the record timeline, not the source.