Forum Replies Created
-
Posts
-
I didn’t know that its just the same worldwide. The things I don’t like:
1. Working at night, because the billionare client watches the spot at 15:00 today, and wants to see his/her revisions tomorrow. The assistants come to you at 17:00, and tell you we have to make the revision for tomorrow. And you work till morning…
2. Making the thing fast!.. When the clients don’t speak eachother, and not talking at the phone, if you work on a shot for more than 10 minutes; some of them start to ask: “Hey, what is your opinion about this shot, how much time will we need to complete this shot?” The more kind ones start to look their watches…
3. Yes, they try to lower the price most of the time.
4. Sometimes they lie to you… They tell that the spot will be broadcasted tomorrow night, so you will have to work till morning. But you see the spot on air, after 2 weeks! And you kindly ask the client, why you had to work till morning for a spot, which had 2 weeks? They lie again, and say that “The boss client went to Paris, and didn’t have the time to approve this, so we waited 2 weeks for him to approve!”… Sometimes agencies lie even to prod. companies like this. But most of the time, the prod. company wants to get rid of the spot, just because they will start a new spot in a few days. That is the major reason…
5. When the directors leave the suite, some of the assistants start to behave like the director. They ask you for things, which will not work. Obviously, it won’t. But if you just tell, it is not a good idea to do so, and if you continue your way, then a cold war starts! You even have to make the assistant of the director happy.
6. Some of younger clients/assistants know a little about desktop compositing softwares. And they start to make suggestions about the techniques of your composition. And they start to tell, “in AfterFX I could make this and this… Isn’t there such a thing in inferno?” And you still have to reply kindly.
Do you know if the 16 Xeon processors are for burn background rendering?
I would also want to see burns using gfx cards for opengl. Burns are really slow in gfx, but CPU operations rock on burn! If all burn nodes had fx4500SLI with each dual core GPUs, inferno would be again the fastest box in the industry, after these upgrades.
pablo_diablo wrote:I’ve tried all combination of parenting, none works…, in manual there is scheme so… it should work, but it doesnt…
where are these example setups? you mean in a form of file?Don’t be afraid, it is very easy 🙂
After parenting your PartGen to your Geometry to draw, just rename your Geometry node to “ParticleDraw”… Remember, the case is important in name. It should be in the manual, but somehow you may have missed it 😉Generally, Discreet uses Seagate drives. So IBM sounds a little bit suspicious 🙁
Once, Discreet sent a self powered extender. And the IBM didn’t have the enough power to extend the port over that length. The system didn’t work for 2-3 days.
At last, Discreet sent an external powered extender, and it worked perfectly! Just be sure that your USB port has the enough power to feed both extender and the tablet! Otherwise, go for an external powered extender.
Xavier wrote:Would exporting your channels as RAW and importing them back convert your motion path to channels?I’m not at the box right now… so I can’t check if it works, but that’s how I “bake” my expressions into curves.
Good luck.
— Xavier
I have tried that in the past. That raw export does export the values in animation channels. So it exports something different, not the axis global position…
This also locks me up, when I want the result values of an animation hierarchy. Let’s say, you have 3 axis connected on top of eachother. You animate all 3 axis, and the axis down at the bottom has some final transformations. There is no way to bake the result of this animation hierarchy into one axis! Altough this sounds theoretically doable, I’ve found no way to do this.
A real life problem to this is: Parent an axis to the camera, and animate the axis, but don’t animate the camera. This is very useful when rotating the cam around the origin. Just imagine you animate the cam on a helical path… When you want to export your camera data as an fbx file. You can not do it! Because your camera doesn’t have any keyframes! 😯
Can anybody think of a solution to this?
I tried to do that some time ago, but “align” only looks to position channel values, and with motion path animation, you don’t have those values! And it “align” doesn’t work with motion path.
Is there any expression, to read that axis coordinate animated along a motion path?
jdboyd wrote:Well, the AJA card at least expects you to already have the frame in 4:2:2 format before it is written to the card. So, if you converted your video to 4:4:4 along the way, then it would be up to you to chroma filter yourself.I think that some of the better GPUs can do the chroma filtering fast enough to work in HD, but I haven’t tried to make it work yet.
A game programmer friend of mine told me, that it is not possible to design a lowpass filter with current generation of GPUs. He said that, this a very computation intensive task, and he talked about curve fitting, which I did not understood too much. But he said that, it is best to design this in hardware. And also I haven’t seen any program on windows, which does software chroma filtering.
jdboyd wrote:I wonder if there would be a market for a YUV composting system.Quantel IQ stores your material in its original format, unless you have to convert it, such as superimposing YUV with RGB material.
jdboyd wrote:A HD-SDI GVO certainly isn’t cheap. But then, perhaps people who can’t afford that could perhaps be happy with analog monitoring (both ATI and Nvidia make a card that offers component HD video out for a second head) or DVI monitoring. I’ve seen a few professional plasma and LCD panels that had both HD-SDI and DVI. Maybe professional displays suitable for fine color work don’t include both yet though.SDI is the expensive thing, and professionals always choose SDI whenever they can. You can have so many problems with analog signals, and that makes SDI so expensive.
jdboyd wrote:v210. Ugh. That is an annoying format. That is the mode I use AJA cards in at work. However, the main path of video through the system stays in v210 format. For some features, we convert portions of a frame at a time from v210 to 16bit (but still keeping the 4:2:2 arangement) and then back to v210.I don’t like v210, altough it saves some space/throughput for quicktime files. If you ask me, decklink should also support other formats, like 16bit per component as you mentioned. At the expense of system bus load, you may free up the CPU.
jdboyd wrote:I wonder if the BMD cards must work in that format though? That format is a standard that you need to use if you are recording to quicktime files. The AJA Xena HS (which is about 50% more money than the Decklink HD) offers 3 RGBA buffer formats for your software to choose from if it is going to insist on processing in RGBA. They are 8bit only though, so perhaps of limited usefulness. You need a Kona2 (I don’t know if any Xena supports this format) to get 10bit RGB.Doesn’t AJA cards use Blackmagic driver? Is Kona2 also like that? If it is so, those cards capabilities shouldn’t be much different…
On 8bit formats, also Decklink supports logical structures like RGBA, UYVY, etc… But not at 10bit! And remember 10bit video I/O precision is very important, even if your material is 8bit RGB. Because at 8bit RGB, you have values: [0..255], but on 8bit YUV; it is [16..235]. So converting to 8bit YUV doesn’t only reduce chroma resolution, but also gives more banding problems. But in 10bit YUV, your range is [64..940] So 10bit YUV is the way to go, for outputting RGB graphics.
jdboyd wrote:So maybe DVS supports some formats that would be useful that the other people don’t, and perhaps the Autodesk software (and most other compositors) benefit from those formats, in which case the DVS cards may be worth the money.If somebody needs something, it may get very expensive. Think the price of a glass of water, for a man dying on desert!.. So if you need the extra features of DVS, you will have to pay, what they like to.
About Decklink, I am very positive. It is a great card for the price, but I just talk about cons/pros.
jdboyd wrote:This could be a bummer, however there are several ways around it. Since it is less than desirable to try to use a Decklink card (or AJA, or DVS) as a GVO anyway when it can be avoided, I don’t see this as being a terribly big problem. The CPU or GPU can do the conversion when writing the file back to disk, if conversion is needed. It seems to me that it is best to try and do as much in YUV space anyway, when working on YUV originated material. I don’t think that this is the Flame way though.When I talked about chroma filtering, I did not mean YUV-RGB conversion. I talked about chroma aliasing problems. That’ all about lowering chroma resolution from 4:4:4 to 4:2:2, and as far as I know, the best way to do it, is to do a lowpass filter to the chroma signal, and this is too much for a CPU, and don’t know if it can be done with a GPU on HD resolution. And yes, you are right, don’t convert from/to YUV, unless you don’t have to. But as far as I know, many compositing apps work in RGB colorspace.
jdboyd wrote:And any video card which will work with a Miranda DVI Ramp (which was what Discreet used for at least some of their PCI-E linux systems prior to the introduction of the fx45000Sdi), and the DVI-Ramp is cheaper than $6k. Other people make converters, but they don’t have as many features as the Miranda. Perhaps they are good enough, I haven’t seen them to judge.I would still go for an fx4500sdi, because the cost of fx4500+miranda DVIramp will be close enough to fx4500sdi. So GVO is not cheap, even on a PC, I did want to mention that.
About the CPU load in Decklink: if you just capture HD, you will have little CPU load like 3-4% The problem comes, when you want to do conversions.
Just look at the below link, and see how to convert it to RGB. Decklink captures 10bit YUV in this structure:
http://developer.apple.com/quicktime/icefloe/dispatch019.html#v210I saw CPU loads like 25-30% on a dual CPU system.
Discreet upgrade policy is always to upgrade to current version. I think its easier to give support only to current version. But it might be possible that, some of the strong clients break these rules sometimes. But in my country, I haven’t heard it. They always push people to upgrade to the latest version.
Now Autodesk pushes people to go under support, or you won’t get minor upgrades!??!!
The price mainly depends on the amount that a board is manufactured. Decklink should be produced in higher amounts when compared to SGI video boards. But not only that, decklink has some missing features…
– Decklink doesn’t support chroma filtering by hardware when converting RGB to YUV4:2:2. And all of the Discreet/SGI hardwares have it, except the O2 video card on older version of flint.
– You have lots of cpu load in capturing, in 10bit formats. SGI handles colorspace conversions by hardware, and they have no cpu load.
– You have GVO (Graphics to Video Option) on SGI, at no computation expense you see your graphics on your broadcast monitor. Decklink is not a GVO, you send the video image seperately to decklink, utilizing your CPU! The only video cards with GVO is, Nvidia fx4000SDI and now fx4500SDI. And they cost US$6k.
– It doesn’t support input and output at the same time. You can not capture video, process it on gfx, and output it with the same graphics card. Remember on onyxRE2 with Sirius video, there were virtual studio apps, which captured incoming video, mapped it on a texture in virtual set, output all the set from sirius video. So you were able to process the video signal. But does AE need this feature? Probably not…
At the end of the day, Decklink is well worth its price. It is a great board. With their new PCIe models, they lowered the costs again in HD workflow. Now those decklinks costs less, and you don’t have to buy expensive motherboards with PCI-X slots. You can build your HD workstation by using your gaming motherboard with PCIe slots.