VideoHelp Forum
+ Reply to Thread
Page 2 of 4
FirstFirst 1 2 3 4 LastLast
Results 31 to 60 of 120
Thread
  1. Originally Posted by jagabo View Post
    The buzzing edges aren't caused by color subsampling (color subsampling makes no difference in grayscale areas). They're caused by insufficient antialiasing and were common in 3d rendering in early 90's. You're stuck with them unless you can increase the output resolution. You might try using a lower antialiasing setting in ACAD and rendering large frames. Maybe you can find a compromise solution that doesn't take too long.
    I tried the same scene @ @2,048 x 1536 in tga It took 5 min and 9.2MB. The buzzing was reduced a good bit at full size, so so after reducing it back to 1,024 x 768

    I'll probably redo the way that I model this particular seam. Right now the black gasket is set back 0.25" like it is in the real world. It leaves a small side edge visible and I think that the high reflective metal finish contributes to the refractions. I'll cheat and bring it out flush. -David
    Quote Quote  
  2. Originally Posted by David Bethel View Post
    I tried the same scene @ @2,048 x 1536 in tga It took 5 min and 9.2MB. The buzzing was reduced a good bit at full size, so so after reducing it back to 1,024 x 768
    Which resize filter did you use? Try Precise Bilinear for the smoothest results (not as sharp as bicubic or lanczos3 but you'll get less buzzing).
    Quote Quote  
  3. I used lanczos3 in Irfanview. I'll look for the Precise Bilinear -David
    Quote Quote  
  4. In VirtaulDub it's called Precise Bilinear. In AviSynth it's just BilinearResize().
    Quote Quote  
  5. I tried to outsmart VirtualDub by resizing to 100 with Precise Bilinear but it was smarter than me. The output files were identical.

    I tried the motion blur and they weren't kidding about BLUR!

    Still digging in. -David
    Quote Quote  
  6. Member
    Join Date
    Nov 2002
    Location
    United States
    Search Comp PM
    Originally Posted by poisondeathray View Post
    Here is an example of an animation, it's slightly smaller 1024x576 @ 30fps ~25Mbps, but encoded with MPEG2 422, with long GOP. Long GOP helps with compression but is more CPU intensive than I-frame only, but I have an single core old laptop that I keep around for testing purposes for "worst case" scenarios (eg. for flash on websites), and it can play MPEG2 422 easily. This can be played on most players eg. smplayer, mpchc, mpc+ffdshow to decode
    I can only play this file in Nero Showtime and VLC. I can open it in Virtualdub and play it but MPC-HC only plays a corrupted looking file. None of my other players will play it including PowerDVD. Is their a setting in MPC-HC I need to use to play this file?
    Quote Quote  
  7. Originally Posted by DarrellS View Post
    Originally Posted by poisondeathray View Post
    Here is an example of an animation, it's slightly smaller 1024x576 @ 30fps ~25Mbps, but encoded with MPEG2 422, with long GOP. Long GOP helps with compression but is more CPU intensive than I-frame only, but I have an single core old laptop that I keep around for testing purposes for "worst case" scenarios (eg. for flash on websites), and it can play MPEG2 422 easily. This can be played on most players eg. smplayer, mpchc, mpc+ffdshow to decode
    I can only play this file in Nero Showtime and VLC. I can open it in Virtualdub and play it but MPC-HC only plays a corrupted looking file. None of my other players will play it including PowerDVD. Is their a setting in MPC-HC I need to use to play this file?
    I just double checked and it's going thru ffdshow

    But I think David has settled on huffyuv, which would be better if his laptop can get realtime playback
    Quote Quote  
  8. Member
    Join Date
    Nov 2002
    Location
    United States
    Search Comp PM
    Originally Posted by poisondeathray View Post
    I just double checked and it's going thru ffdshow

    But I think David has settled on huffyuv, which would be better if his laptop can get realtime playback
    Would Lagarith be a better option? I've heard it's faster than Huffyuv and compresses better.


    I have a much faster system but my way of creating animations usually starts with Photoshop 5.5 where I create a PSD file from all my graphics files. I then open the PSD file in Gif Movie Gear where I either save as a SWF file or uncompressed AVI that I open in Virtualdub, alter the framerate and save as compressed AVI.

    I'm not working with anything like either of you are though.

    I used to have MGI Video Wave 4 which would allow you to add graphics files and had an option for smoothing out the transitions between frames. Back then, I was using 98SE on a Celeron and W2000 on a 3.2 P4.
    Quote Quote  
  9. Would Lagarith be a better option? I've heard it's faster than Huffyuv and compresses better.
    Lagarith compresses better, but decodes much slower (about 6-8x slower than UT on a quad core)

    UT is the fastest lossless codec on multicore systems (it's almost 1.5x faster than huffyuv-mt for decoding speed)

    @David - avisynth has some options for supersampling and antialiasing as well that might help. Can you post a .png sample of that frame? or even a frame in which you've rendered to a larger dimension ?

    motion blur helps more when there is motion between frames (like faster camera movements or object movements), not so much for edge aliasing, as is the case for you
    Quote Quote  
  10. In the video world, what does faster refer to? Compiling speed? CPU play back speed? Decoding speed?

    In the CAD world it is strictly about CPU number crunching. Process the data and then display the results. -David
    Quote Quote  
  11. I was referring to decode speed (i.e. playback) - which is important given your plans to playback for an audience on a laptop. If you can't display all your hard work or playback is choppy, that doesn't do any good.

    Decode speed is more important for video editing in NLE's like Premiere Pro , vegas, etc.. where timing is important for transitions, syncing to audio, etc...

    But when discussing lossless codecs, there is also encode speed, cpu usage, compression to consider

    Formats that have higher compression usually require more cpu usage to decode (because it has to "uncompress" the data). Most lossless codecs have selectable options for either better decode speed or better compression

    Uncompressed formats have large data rates, and therefore storage transfer speeds come into play. You can't playback an uncompressed 1080p24 stream with a single mechanical HDD in realtime. You need at least 2 in RAID0. Higher color sampling and bit depth also require more bandwidth. That is why earlier we were looking at other alternative lossy formats
    Last edited by poisondeathray; 3rd Mar 2010 at 15:39.
    Quote Quote  
  12. Heres the png zipped up. Still 1 MB for 2048 x 1536


    I'm used 2,000 frames / hour to compile. So everything I found here is extremely fast. 3,000 in under 5 minutes YES!


    I'm begin to understand the decoding speed importance. -David
    Image Attached Files
    Quote Quote  
  13. I had a look at it with some filters, unfortunately many of the video filters work in different colorspace (YUV), and you don't want to do colorspace conversions back & forth (it's lossy)

    I can't think of anything else to help with the aliasing post production or after the fact (short of a destructive blur); I think it has to be done within the original program.

    You could rotoscope or apply filters to sections (so as not to damage the rest of the "good" parts), but this is tedious to do
    Quote Quote  
  14. Avisynth looks fairly straight forward. Just learn a bit of language. I code a lot in AutoLISP. I wrote the walkthru code a long time back and could incorptrate their script right into the program. This is a snippet of making the rendered image


    Code:
                    ((= stype "AR2")
                     (ar_force_translation)
                     (ar_render 2 0 0 1024 768)
                     (ar_jpeg_compression_level 1)
                     (ar_save_image (strcat "c:/tga/" pre
                                                  (cond ((< i 10) "0000")
                                                        ((< i 100) "000")
                                                        ((< i 1000) "00")
                                                        ((< i 10000) "0")
                                                        (T            ""))
                                                   (rtos i 2 0)
                                                   ;;; ".TGA"))
                                                   ".JPG") 24)
                     ))
    So by simply change file file extension from ".JPG" to ".TGA", Accurender knows how to output the frame.

    Adding file info or counters would be easy to do. -David
    Quote Quote  
  15. Largaith was very fast and VERY jerky.

    Cinepak by Radius is looking pretty good too.

    I've dragging and dropping the files into VirtualDub. Is there an import or multi file open thingy that I'm missing? -David
    Quote Quote  
  16. Originally Posted by David Bethel View Post
    I tried to outsmart VirtualDub by resizing to 100 with Precise Bilinear but it was smarter than me. The output files were identical.
    Of course. Here's your 2048x1536 frame resized to 1024x768 with VirtualDub's Bilinear option:

    Click image for larger version

Name:	1024.png
Views:	183
Size:	517.0 KB
ID:	655

    Lossless codecs like Lagarith and HuffYUV aren't meant for playback. They're for capture or intermediate storage. Cinepak is a 20 year old codec that has no place in modern video.

    VirtualDub: File -> Open Video File -> pick the first image, it will automatically append the rest (if they're sequentially numbered).
    Last edited by jagabo; 3rd Mar 2010 at 16:20.
    Quote Quote  
  17. Originally Posted by David Bethel View Post
    Largaith was very fast and VERY jerky.
    I should clarify what is meant by "fast" in video terms.

    It implies the highest fps that you can play without dropping frames. If you have to drop frames because your cpu can't render them fast enough, you get jerky playback. In video terms, this is "slow" for playback

    A higher minimum fps playback, if above your video's fps, means realtime playback. In your case 30 frames per second. If you could only play 29fps, you would have 1 dropped frame every 30 frame cycle

    If your minimum fps is like 250, you are never in danger of dropping frames, and always get the desired playback. You have cpu cycles left over to do other stuff, bake a cake, etc... Note just because you can play 250fps, it doesn't speed up your video like "fast forward", it only plays at the desired 30fps, but it's nice to know that you have extra headroom to play with




    I've dragging and dropping the files into VirtualDub. Is there an import or multi file open thingy that I'm missing? -David

    Nope , whatever works
    Quote Quote  
  18. First off, let me thank everyone for very fast and courteous responses to a newbie on a subject that has probably been covered a multitude of times, but that I could not find in the forum search.

    If your minimum fps is like 250, you are never in danger of dropping frames, and always get the desired playback
    Now that be great bench mark test:
    • Run a test and the app tells you that your system can handle nnn fps using this codec or this media player!

    Maybe already out there but I've never seen it.


    I'll run half of animation in tga starting tonight -David
    Quote Quote  
  19. Originally Posted by jagabo View Post
    Here's your 2048x1536 frame resized to 1024x768 with VirtualDub's Bilinear option:
    I got about they same results on this end with VirtualDub so I must be doing something correct. It just adds a ton of time on the front end.

    I did one project a few years ago that took 600 hrs of CPU time. I made a render farm here in my office but wasn't really that impressed with the splicing due to the differences in the machines. You could definitely tell when one machine stopped and another picked up. I think it was the differences in the video cards on the machines. Everything else ( cad package, rendering software, 3d models ) were exact duplicates. -David
    Quote Quote  
  20. Originally Posted by David Bethel View Post
    I did one project a few years ago that took 600 hrs of CPU time. I made a render farm here in my office but wasn't really that impressed with the splicing due to the differences in the machines. You could definitely tell when one machine stopped and another picked up. I think it was the differences in the video cards on the machines. Everything else ( cad package, rendering software, 3d models ) were exact duplicates.
    I don't know about Autocad, but in After Effects (not 3d rendering, but 2.5d), each frame can be rendered independent of others, so you get near perfect scaling with a render farm, even with different speed computers and setups. Each starts up in a different segment location in the sequence, so it truly is "parallel" rendering
    Quote Quote  
  21. Originally Posted by poisondeathray View Post
    I don't know about Autocad, but in After Effects (not 3d rendering, but 2.5d), each frame can be rendered independent of others, so you get near perfect scaling with a render farm, even with different speed computers and setups. Each starts up in a different segment location in the sequence, so it truly is "parallel" rendering
    In AutoCAD, you use the same premise. You could have 2 machines doing every other frame. The render package moves the camera and target along spline paths, sets the distance & lense to pre determined values, etc , renders the image, then deals with local machine to figure out what to do when saving the image file. I added right and left panning to mine just so you get a bit of different viewing angles of the equipment.

    I know Adobe makes good products. I'm just not a fan of their GUIs. I'm still a command line / OS guy at heart. -David
    Quote Quote  
  22. Originally Posted by David Bethel View Post
    I got about they same results on this end with VirtualDub so I must be doing something correct. It just adds a ton of time on the front end.
    I think you are missing the point. Look at the thinnest lines in the downsized video compared to similar with lines in your original videos. The new ones are much smoother:

    Name:  comp.png
Views: 292
Size:  16.8 KB

    Those rough broken lines are causing the buzzing edges you see during motion.

    Over-rendering like this is one of the basic methods of antialiasing.
    Last edited by jagabo; 3rd Mar 2010 at 20:23.
    Quote Quote  
  23. Originally Posted by jagabo View Post
    I think you are missing the point Over-rendering like this is one of the basic methods of antialiasing.
    I understand. I could see the difference of line edges on the frame that I did here @ 2,048 x 1,538 and then did the resizing. It's just that it will take 3.5 times longer to render the images. Maybe I'll try 1,536 x 1,152 ( 50% bigger ) and see how long that takes. This particular presentation is (2) 1,514 frame animations spliced together, once with a forward path, once reversed. 3028 total. -David
    Quote Quote  
  24. OK!

    Got the rendering done in TGA @1,024 x 768 3.74GB

    Compressed using huffyuv 2.1.1 in VirtualDub 1.9.8 to 1.10GB, very nice, very very fast ( under 3 minutes )

    The black seams are definitely modelling problems on my end.

    I think that the most of the remaining jerkiness is coming from rotating the camera too quickly.

    I'll try some other codecs as see how they go.

    My laptop HD led is flickering away, but it's not on steady and I'm not dropping frames, so I probably have a little more performance to squeeze out of it. Amazingly it still plays the stuff back better than my desktop machines. I do keep it stripped down as far as running many background services etc. Thanks a ton! -David
    Quote Quote  
  25. 30 fps is jerky. The jerks aer fast enough that many people would say it flickers. If your computer can play 60 fps video fast enough see the video in this post comparing 24, 30 and 60 fps:

    https://forum.videohelp.com/threads/307004-Best-framerate-conversion-%28eg-23-97-to-30-...=1#post1888926
    Quote Quote  
  26. 1,514 frames @ 30 fps 50.4 seconds

    This is strange. Compile avi @ 60fps using huff ( changed both the option -> preference -> image and video -> frame rate settings and restarted VirtualDub )

    smplayer and windows media 11 says it is 25 seconds long and the progress bar shows the same. But it takes 50 seconds to run.

    Irfanview tried to play it 12 seconds ( way fast and way ugly )

    Could I have forced a skip no frame play back option somewhere and this is a fast as players can handle? -David
    Quote Quote  
  27. Even weirder. I had never actually time time playback.

    One desktop takes 85 seconds, the laptop 76. 19 fps WOW! No wonder they look jerky at times.


    The black circles makes a very good demonstration. -David
    Quote Quote  
  28. I don't understand; are you saying that your are dropping frames and can't get realtime playback? (not surprising for your laptop)

    To double check, encode to something easy to decode , like xvid and compare

    If your composition was timed for 30fps, you should encode it at 30fps.
    Last edited by poisondeathray; 5th Mar 2010 at 09:43.
    Quote Quote  
  29. It really doesn't look like it is dropping frames.

    I've been making the avis and testing them on a AMD 64x2 Dual, 2GB Ram, GeForce 6150. It has been a tick slower than the laptop.

    I encoded using xvid. 30 fps, playback time 50 seconds media player 11. Maybe it's huff ???? -David
    Quote Quote  
  30. As stated earlier, lossless compression is not meant for playback. UT is slightly better than huffyuv for multicore computers (about 1.5x faster decode speed), but on a single core I doubt you will get realtime playback. Forget about lagarith, impossible to get real time.

    This is why earlier on we were looking at other forms of compression that were easier to decode, yet were not 4:2:0 sampling.

    It's probably 50s long, and your 60fps version is 25s long. Since your 60fps version took 50s you're dropping frames. If it plays too fast , like 12s in irfanview, you're still dropping frames.

    The xvid version is what you want it to look like in terms of playback speed, but unforunately it's 4:2:0. It may or may not be "good enough" for you
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!