[Sinan]

I miss the tips/tricks section John. Everybody was contributing there. And/or you were adding tricks from flame-news. But that section is not updated for very long time. No tips for long time?..

[Sinan]

@julik 26315 wrote:

Is there an expression which will allow me to fade between two animations? Let’s say I have camera motion in camera A up to frame 100, and camera motion in camera B starting from frame 80. I would like to make a camera that would smoothly transition between the animation of A and B across frames 80 to 100. Something like constraints/Trax in Maya.

Did anyone come up with something similar for flame? maybe there is a spark which just mixes curves and allows you to copy to it and back?

I made that camera transition thing for 2 x 3D tracked cameras. All I did was creating a 3rd camera, and a controller axis. I wrote something like (sorry I am not at the box right now):

camera3=”camera1*(1-controller.position.x)+camera2*(controller.position.x)”

And I animated my controller axis from 0 to 1, to animate between cam1 & cam2.

hope this helps,

[Sinan]

@cpage 26275 wrote:

Hi all,
I know this question turn around the world a lot of time but i’m sorry to say that,
i never find the answer 😮
Maybe it was right in the past that Inferno worked with 3D tracker, multi CPU and real time
but it’s also the case now for Flame (even in 2K).
So please, someone will be helpful to give me (us) the real good differences.
Thank you in advance.
Cp

The quick answer is flame2009=inferno2009 except the startup screen / message bar / cursor shape / executable name / user name… Did I forget something? 🙂

But! Or even a big BUT! Most of our clients still think that: inferno>flame. Especially in Europe, it is like that. Remember how fast an onyx2 renders compared to oct1mx? This is history now, but Autodesk couldn’t kill inferno licenses. Because we, the users who paid more for the software license, don’t want to downgrade their suite names to flame!

But from an engineering perspective, there is no use to have a product named “inferno” for today. From sales point of view, inferno is a big brand. I suspect that, AME might do something with the inferno name in the future. For example extra software features??! Who knows…

In the past, inferno has meant speed most of the time. Although the first versions of inferno had 12bit feature, which flame didn’t have. And that was the main difference, making inferno capable to do 10bit log film work. And it shared the same onyx1 platform with flame. But very very old versions… Now again inferno and flame run on the same HP box. Today it is a HPxw8600/8core/nvQ5600sdi, now inferno=flame. But today, inferno doesn’t have any feature difference!

[Sinan]

@Unkle 26193 wrote:

If your footage is 16:9FHA then you need letterbox it in action, i’m not infront of the box right now so i’m not to sure. But once you’ve done this you’ll want to

1: timewarp your footage to double the length of the spot if it were running at NTSC. So if it’s a 30 second spot, timewarp it to 1800 frames
2: Then interlace it, you will be left with a 900 frame clip
3: Then using resize. Resize it 720 x 486, play around with the crops whatever looks best, a letterbox should work well.

I would first resize, then TW the sequence This will provide better scaling and interlacing…

Another idea, if you can make spot 4% longer in time, then just create a 24fps sequence, then add pulldown, and TW audio. If you don’t want to have TW artifacts. Ofcourse this can be lots of trouble to the client, since it will change the length of spot.

[Sinan]

Maybe it is a “one frame optimization” thing. Remember for one frame media, if it is not animated, an action node calculates CC, blur, key, only for the first frame. Do you have lots of single frame media in your action nodes? If it is like that, all action nodes would cache all media processing for the first frame, and then the next frames will go a lot faster. But 5 min. for an initial start seems to be a huge delay. However it might be possible if you have a huge setup with lots of action nodes, and media processing…

[Sinan]

I don’t think that Chris wants to make advertising for Filmlight. He just wants to tell his opinion about Truelight. Maybe he is too excited about Resolve, and that Baselight is faster. Hey, who needs more than 20 secondaries at 4K anyway? Resolve is already doing a pretty good job! I guess that, Chris was excited about Baselight optimizations, which gives more performance then Resolve. When I saw the first flint/linux box, people might have thought that I am advertising Discreet! Hey, I never worked for them. But I am sure that I have written lots of excited posts about the Discreet linux boxes.

And one more thing… If you ask me, I will tell you that RSR Cinespace is the system to go for a lot of reasons. But these are my opinions, and it wouldn’t mean that I am advertising Cinespace. Cinespace lut formats are open, you can export them to any format. You can use most of the industry standart probes, and they don’t sell you a Cinespace probe. But I never used Truelight, and can not tell about cons/pros, or compare Cinespace to Truelight…

Ofcourse Chris is a special case, since he worked for Filmlight. But I guess that he is back to flame world for the last couple of months. And some of the newer emails are, after he came back to flame world.

About his Red post, I totally agree that Red doesn’t have a stunning image quality. It is no magic! It is close to 35mm sized sensor, and if you ask me, it shoots video like images with a bigger sensor. I would expect that Arri D21 would provide better quality, and the same sensor size as a 35mm neg. frame… But this doesn’t make Red a bad solution. It needs little storage, uses cheap recording devices. No uncompressed huge raids on set! It is a cheaper digital cinema solution. Small, light, affordable compared to many other solutions…

Did I make any advertisements for any brands here?

[Sinan]

@Alfafa 24768 wrote:

Hi everybody,

(I’m using Flame 2007 on Linux)

Does anyone knows how the camera projection works ?

I had to remove some wires on a shot with a huge nodal pan, so a lot a parallax…

I 3D tracked it fine, then the idea was a to clean a frame and project it in scene, but I could NOT get it to work at all.

If someone can explain me the use of that Projector thing that would be terrific and make my day !

Thank you youyou.

Alfafa

Is it a nodal pan? If it is very nodal, then you shouldn’t have any perspective changes.

[Sinan]

If you have many action nodes, your rendering speed will be a lot slower than your local render. Especially if you are multisampling, you will see much more speed difference. The reason for that is, current burn version uses software opengl, so no gfx card texturing. But if you have a cpu type processing batch tree, all that matters should be the network transfer speed. You should be able to get at least 50% of your local machine speed on cpu type setups.

But if you use optics, you will see that burn is way slower!

[Sinan]

This sounds very weird to me. But I’ve seen some bugs with userfun expressions. Instead I write my own expression, and don’t use them, just because of bugs. However these impressions are from inferno6.2, now we are on 2008 and I didn’t have time to test if the bugs are still there. So I assume you have the same old problem with userfun expressions, then I would try this:

media1.matte.slip = MySlipAxis.position.x – frame

This way you get rid of frametoslip userfun expression, and here is what happens. At frame#1, you want to display text#1, and you set a constant keyframe to frame#1 of action, until the next cut arrives at frame#38. Example:

frame#1, position.x=1 then slip = 0
frame#15, position.x=1 then slip = -14

Since the slip=-14, action is at frame#15, the first frame of matte clip will be repeated for 14 frames! When we arrive to the next cut point at frame#38, just set text#=23. Set the position.x to 23.

At frame#38:
frame#38, position.x=23 then slip = -15, so the clip starts playin back at frame#16, and when we arrive to frame#38, 23 frames has been passed, so it will show frame#23 on the matte clip.
___________________________________________

Hope this helps, and I guess this is the same problem happened to me years ago. I used this method, and everytime I loaded my action setups, they worked w/o any problems.

[Sinan]

@JBullard 24374 wrote:

Wire is not available for Irix systems. Is that correct?

We have an onyx350 running inferno2008, and it links very well to our linux inferno/smoke_2008 boxes via gigabit ethernet. Maybe you talk about infiniband support on irix. You can’t have infiniband on irix, which is a lot faster than gigabit.

[Sinan]

@brianH 24216 wrote:

If I’m doing heavy compositing, I usually go GPU-bound before I go CPU-bound. I’ve watched 45-minute HD batch renders sit at around 55% CPU utilization on my HP machine. C’mon Nvidia! 😉

It might be because of the slower readback of gfx cards. Does anybody know if linux boxes use hardware accumulation buffer for antialiasing samples?

[Sinan]

Flame 2007 to Flame 2008 upgrade is a very big major release. I have seen the beta at IBC07, it is a great jump. If you look to AME teams development speed for the past few years, you would expect flame 2008 features, maybe at year 2010. But they are faster this year somehow, I can say it is worth something like 3 major versions upgrade from 2007 to 2008 🙂

I can’t wait to see what they can do for 2009 release, after dropping irix support. I expect AME to make use of pixel shaders v4, and make the PC platform even faster for some of the current CPU based operations (like CC, blur, keyers, etc..), all of these operations might get lots of boost with a GPU implementation.

[Sinan]

@Mon1018 24127 wrote:

i only know that it is still available for IRIX on Tezro

Perhaps 2008 will be the last version that will run on tezro, and we should be happy about this. I hope then, AME will move to new opengl techniques like pixel shaders. Dropping the old sgi platform will open new room for improvement in software.

[Sinan]

@vincefaro 24120 wrote:

I want to use a boujou camera to generate motion blur on a clip in action. Is this possible without the image actually moving, ie the old trick where you attach a node to a clip to generate motion blur. Can this be done with a camera?
Thanks

the new motionblur node in flame 2008 is able to do what you need. But the old trick you ask for, is only in X-Y axis, and the same blur applies to the whole image 🙁

However, there are some solutions with current software. You can try an optical flow timewarp to slowdown, then compound the resulting clip a few frames, you might get some motionblur with optical flow.

But I don’t see an easy way of doing this in action. Unless you make a rough model, and project your footage to that model, and render your boujou camera with moblur. It seems to be possible, but that would be a lot of work!

[Sinan]

@NumbThumb 24113 wrote:

Not too sure about the first bit, but if you wanted to displace both the R & B channels at the same time you could use two mono nodes (one for each color) and then screen or add them together and then use that as your displacement matte.

I don’t think that this would be possible in one pass using only one image node. You should first displace an image in X direction, using your R input, then displace the result again in Y, using B channel. Then you can blend R&B channels with any method you like, and produce a new channel. Then use that new channel, as the displacements source for Z direction.

Methods may differ:
1. you might render clips, process the passes.
2. you might use batch, and each pass would be an action node
3. use the source nodes, to pass the results within the same action module (this should be the fastest one)

[Author]

Posts