CompuServe Thread

#3DS Flic vs NumPic

21 messages in this thread
#141302From: Brent BonetDec 12, 1994 8:49 AM
I've got an interesting question. I've been working on an animation in 3D Studio R3 and when I render the animation as a flic (100 frames at 512×486) the flic ends up being in the 5 MB range. As I got closer to the final render I decided to render TGA's and convert them to 8 bit BMPs and import them into Animator Pro using numpic to make the flic. The flic size is about 21 MB!!! Why is this? Brent Bonet – POP! Multimedia Atlanta
#141369From: Nik Grant [adesk]Dec 12, 1994 1:18 PM
The size is quite different, agreed. I would guess that the palette and image is what is causing the differences. 3DS is capable of some pretty subtle placement of pixels. If you have a background that is not plain black you might find that its "description" varies considerably from frame to frame. Outputting to FLIC from 3DS optimizes the colors differently than your AAPro route to a flic. There may be other responses from people who have had some similar experience who can confirm this.
#141393From: NAAUGDec 12, 1994 2:32 PM
/* I've got an interesting question. I've been working on an animation in 3D Studio R3 and when I render the animation as a flic (100 frames at 512×486) the flic ends up being in the 5 MB range. As I got closer to the final render I decided to render TGA's and convert them to 8 bit BMPs and import them into Animator Pro using numpic to make the flic. The flic size is about 21 MB!!! Why is this? Brent Bonet – POP! Multimedia Atlanta */ Before I take a guess, could you try this: use ANICONV.EXE and import the 3DS flic, save it out as an 8-bit flic. What size is it now? -dave
#141410From: Brent BonetDec 12, 1994 3:37 PM
Aniconv saves it at the same size…5MB. Brent Bonet – POP! Multimedia Atlanta
#141470From: NAAUGDec 12, 1994 8:01 PM
/* Aniconv saves it at the same size…5MB. */ Wow. Well, I guess there goes that theory. <g> I know that there are different types of BMP files (256 is just one flavor), thought perhaps FLC file had 24-bit info in it (contributing to filesize). If you've processed the 21MB with ANICONV into an 8-bitter, and it's back down to 5MB, something major had to give. When you configured 3DS for BMPs, did you configure for 256 BMP or "True" BMP (there is a significant difference)? -dave
#141535From: Brent BonetDec 13, 1994 8:17 AM
I'm saving in TGA then I use Hijaak to convert to 8 bit BMPs. I then use numpic to import and a use one palette to apply the same palette to all frames. The idea is that when I get close to a final product, to render and create a flic. If it's what the client wants I can then just record to CRV without having to re-render the file that I just spent 10 hours rendering! Thanks, Brent Bonet – POP! Multimedia Atlanta
#141583From: Yost GroupDec 13, 1994 11:10 AM
Here's a guess (don't have the time to test it out). When you create your animation by outputting separate 24-bit TGA images, each image has a different random dithering, whereas in flic output the dithering is controlled for the flic in such a way that you don't get "dither-crawl" during playback. As a result, the deltas for a flic generated directly from 3DS are much smaller than those in a flic that's been converted from a series of TGA's. In the converted flic, because of the dithering, probably every pixel changes in every frame — resulting in a gigantic flic. – Jack
#141659From: NAAUGDec 13, 1994 6:40 PM
/* I'm saving in TGA then I use Hijaak to convert to 8 bit BMPs. I then use numpic to import and a use one palette to apply the same palette to all frames. The idea is that when I get close to a final product, to render and create a flic. If it's what the client wants I can then just record to CRV without having to re-render the file that I just spent 10 hours rendering! Thanks, Brent Bonet – POP! Multimedia Atlanta */ Welp, there's only so many things that can cause this. My guess Brent would be that by using individual TGAs, no "squeezing" is being done on the "near colors" between adjacent frames. Consequently, each frame is "on it's own", nothing to offer in repetition for the next frame. I tried a few things last night to verify this… I rendered out to individual TGAs in 3DS, then used NUMPIC on them to build a flic. I also generated a flic rendering out to an FLC from 3DS. I got the same results. NUMPIC is simply building a frame-by-frame flic by appending each new frame's palette and pixel values to the last treated frame. It isn't optimizing as it goes along appending. The TGA's palettes are so huge that the likelihood of two frames having the same exact color values is practically nil if cameras and lights are in motion. It's only after the entire flic has been constructed from the radically different individual images that palette optimization functions can be applied to it and you'll start to get some repeating palette values between adjacent frames. If you want individual frames, why not generate the flic using High Palette mode in 3DS and then use NUMPIC to bust it up into individual frames (images) quickly rather than re-rendering? Am I missing something here? <g> -dave
#141672From: Brent BonetDec 13, 1994 7:44 PM
It would be really nice (and time saving) if you could apply the palette processing that is done after rendering a flic in 3DS to a series of TGA files. It seems like it would be possible. I'm just looking for an efficient way to present my animations before recording to tape or CRV. I have to borrow a CRV to render so it's much easier to present a FLC for client approval. Brent Bonet – POP! Multimedia Atlanta
#141692From: NAAUGDec 13, 1994 10:28 PM
/* It would be really nice (and time saving) if you could apply the palette processing that is done after rendering a flic in 3DS to a series of TGA files. It seems like it would be possible. I'm just looking for an efficient way to present my animations before recording to tape or CRV. I have to borrow a CRV to render so it's much easier to present a FLC for client approval. Brent Bonet – POP! Multimedia Atlanta */ Maybe ANI isn't the answer with this case. Have you tried posting this same task in section #1? Perhaps we're missing something obvious on this one. Normally I'd generate the flic in 3DS, and then bust it out into individual images if needed with ANI's NUMPIC. -dave
#141694From: Nancy Fulton, OpenCADDec 13, 1994 10:43 PM
Hi! Why do this in Ani Pro (a fine product goodness knows). Just stuff a list of TGA file names in an ASCII text file. Put the TGA Files and the text file (which should have the extension .IFL (mytgas.IFL)) in a directory in your map paths (C:\3DS4\MAPS). Then select Renderer/video Post from the Keyframer. Enter the total number of frames you want to combine in the field next to the word SET at the top of the screen. Then select SET to create that number of frames. Select Add and click in area under the word QUEUE. Select the edit button. Click on the [KF SCENE] entry. Select the bitmap button and then identify your IFL file. Then exit back to the Video Post main dialog box. And select the Render button. This will convert all your TGA's to a FLIC inside 3DS (and will generate a single pallete as a by product. Furthermore – you can also convert all your TGA's into BMPs Gifs or any other 3DS supported format using the same technique – just select the Configure button in the Rendering dialog box to specify the output file. Learning this much video post takes less than half an hour and will save you hundreds of hours of rendering time. (Not that I don't love Ani Pro – I use it all the time to create custom pallets….) Seriously, its worth the hassle. At any rate – Sorry to butt in but I thought you might like to know about this solution if you didn't already which you might. Its certainly easier than HiJack Meets AniPro. Cool, Nancy Fulton (PS: You can also select that HIGH pallette option which should help reduce screen crawl if such a thing is really caused by a screwed up pallett rather than just a sincere limitation caused by having only 256 colors to work with).
#141743From: Brent BonetDec 14, 1994 6:36 AM
Nancy! That's probably the answer! I'll give it a shot! I haven't gotten too heavily into video post yet, so… Thanks, Brent Bonet – POP! Multimedia
#142095From: Nancy Fulton, OpenCADDec 15, 1994 9:58 PM
Dear Brent: You must learn Video Post. It is amazing. And another thing one of my students just taught me. You can render animations in reverse by using the Ranges option in the render dialog box. Switch the Starting and Ending numbers – instead of 1 to 30, render 30 to 1 and you get the animation backwards. This can be so HANDY for those repetitive animations…. Cool? N. PS: I know Video Post will do your job for you. Let me know if you need more help.
#141749From: Ted BoardmanDec 14, 1994 7:28 AM
Hi Nancy, Good to see you around the neighborhood again! Are you back home? Later TedB !^NavFont01F0007MGHHG4800A9
#142102From: Nancy Fulton, OpenCADDec 15, 1994 10:03 PM
ZZTed <g> I hope you are very well and yes I'm back. I finished that book so I'm only completely overworked and without a moment to myself rather than ridiculously overworked without a moment to myself. Are you well? Teaching many classes? OpenCAD Incorporated (yeah, we're a corporation now) just did 170 hours of training on 3DS in the last 2 months – 3 classes for UCLA and one custom class. Sounds like YOUR schedule eh? Fortunately I'm off for the rest of the quarter – nothing till Feb of next year. Man am I tired. Stay warm and Merry Xmas Ted. Take time off for the holidays…. Love N
#142140From: Ted BoardmanDec 16, 1994 5:14 AM
Nancy Geez, 170 hrs of training in two months, brutal! My schedule is nothing like that thank you, but I am managing to stay plenty busy<g> Ken Flesher has just built a beautiful new lab at UNH so it's a pleasure to do classes there. I've got Acad, Designer, 3DS, and AutoArchitect classes scheduled for this winter, plus, plus, so I guess I won't get "bored, man"<g> Hi to Mark and the baby! TedB !^NavFont01F0007MGHHH45FD35
#142649From: Nancy Fulton, OpenCADDec 19, 1994 3:07 PM
Dear Ted: Ah AutoArchitect. Its been so LONG since I taught that…. I'm doing an AutoCAD class inf february I think. And I'm writing about Designer but not teaching it. I admire you for the diversity of your curriculum. Its hard to stay current on so many applications. Best wishes for Xmas and 1995, and I hope to see you before the next year ends (is it too early to ask about Siggraph?<g>) N.
#142782From: Ted BoardmanDec 20, 1994 7:40 AM
Nancy Diversity has it's downside, too. Can't be good at any one.<G> I've pretty much phased out any real architectural jobs, so if they make any conceptual changes to AA I probably won't be able to keep up. Yeah, let's wait until after the new year to talk about Siggraph. Thanks TedB !^NavFont01F0007MGHHGEBE41C
#141501From: Steven ElliottDec 12, 1994 11:01 PM
Hi Brent, I'm taking a WAG here… Rendering a flic in 3DS always ends up with a single palette used by all of the frames in the flic. When you rendered to TGA's, each frame was 24-bit (no palette, 16 mil colors). Converting to 8-bit BMP's created 100 BMP's, each with their own 256 color palette. Numpic loaded those BMP's with no palette conversion so you ended up with a 100 frame flic where each frame had a different palette. This essentially destroyed any change of getting decent flic compression. Big, BIG, flic file :). Processing with ANICONV (or using One Palette after Numpic) normalizes all of the frames to use the same palette. This gives you much greater flic file compression. Steven
#141534From: Brent BonetDec 13, 1994 8:17 AM
Hi Steve!! Nope…not right. I have converted with with one palette and although I didn't save the file until I applied a single palette, the file size is still HUGE! When I get close to a final product I don't want to render once to see a flic and then once to get the final TGAs. 21MB for 100 frames doesn't quite playback well on my PC! This is rather disturbing. Thanks, Brent Bonet – POP! Multimedia Atlanta
#141565From: Nik Grant [adesk]Dec 13, 1994 9:54 AM
What must make it big is the delta – the size difference of each frame reported in the Extra/Info dialog box. Sorry that we're all guessing here as to what is going on. My guess is that you have a moving camera in your animation (vs. a plain background with a small amount of movement local to one area the way that the sample UNIONSTA.FLC works). This would cause the background to be rendered differently for each frame for both the straight flic generation and the targa generation. However, I would guess some logic during the flic creation would minimize/optimize these differences better than when assembling 100 "unrelated" files.