| Author | Message | ||
     Robert Hofmeyr Rating: N/A |
Have been browsing this forum for a while and have found a lot of useful info. So thanks to everyone for their time. We are trying to set up a low-budget stereo video solution in Cape Town, South Africa. Our current plan is to purchase 2 miniDV camcorders (not genlocked) to shoot the videos then capture the left and right views onto our Final Cut Pro editing system. We will then manually sync the views, lock them together and edit the video (obviously, we will only be able to view either the left or right view while editing). Once the edit is complete, we will export separate left and right videos for projection. We are not keen to convert to interlaced stereo video as that will result in quality loss and we want to maintain a full quality non-stereo version. Also, if we convert to field-sequential video, we will not be able to preview while we are editing. However, it would be nice to be able to create a field-sequential video once editing is complete. What is the best not-too-expensive software tool for combining left and right DV video into field-sequential 3D? We would prefer to use macs but have PCs as well. Assuming our final product is 2 separate left and right DV videos, what is the easiest way to sync playback to 2 projectors? Getting 2 miniDV decks to sync will be a nightmare. If we were to put both videos onto a PC, and bought a good graphics card, is there playback software that will output 2 streams from different files in sync? What graphics card would you suggest for this purpose (we have a P4 running XP)? Does this seem like a crazy way to create stereo video? Any other suggestions are welcome. | ||
| Puppet Kite Kid Rating: N/A |
Great questions. It's always good to see more people get interested in this. There is a very active group of people doing this in a Yahoo group who have already solved many of your challenges: http://groups.yahoo.com/group/3dtv/ There are many things to cover, and the best way is to start from the beginning and work through the process, depending on your equipment, software and needs. Important things are: Do your cams have LANC? If so, they can be fairly accurately synced while shooting. If not, there are a couple of ways to monitor your degree of sync. There are also *advanced* ways to sync during editing to the nearest field. There are freeware solutions for the PC, and in my opinion, Adobe After Effects is the best overall commercial software for 3D compositing (PC or Mac), as you can drag both videos side-by-side into one composition window, separate fields, do all the needed compositional changes, etc... all quite easily. There are also some existing PC freeware solutions for playback. Most of use use large parallel videos on fast computers, as there is no need for additional syncing or genlocking. A couple of freeware players have dual output capability. I could go on and on... but it's best to address each issue one at a time. Feel free to ask more specific questions and/or join the Yahoo group. P. K. Kid Non-commercial stereoscopic 3D video: (All G-Rated) http://www.PuppetKites.net | ||
| Robert Hofmeyr Rating: N/A |
Thanks. I had already been to your site. There is a lot of useful information there and it made me realise that it is possible to produce stereo video on a limited budget (most of the other sites with good info are also trying to sell some expensive gear). Yahoo group looks interesting. Shall take a look this evening. Cheers. | ||
| M.H. Rating: N/A |
Robert: During using of 2 digital DV camera you will fall in several non-avoidalble problems witch will totaly degradate your work: - the cameras are not genclocked, you will be not able to record anything with serious motion - the camera record iterlaced, image will be cripled by deinterlacing artefacts - zoom and optic on this to camera are not synchronized , it is amlost impossible to use zoom and brightness/contrast are non-synchronized as well ... Several people are trying to solve this problems by software post-processing tools, but the results are always pure in comparison with other methods. | ||
| Puppet Kite Kid Rating: N/A |
We now have a couple of good solutions for syncing the cams. For example, Rob Crockett has made some tools for syncing with LANC and/or monitoring the degree of sync (plans of some of them are also on his webpages). For example, I am now using two cams _without_ LANC that are monitored with a tool (called an "LCD Sync Shepherd"). For example, I can sustain 1/500th of a second sync or better for up to 20 minutes until my cams go out of sync more than 1/500th of a second. The LANC controller, if you have LANC, can power up the cams simultaneously with a very high degree of sync, that is sustained for the amount of time until the cams go out of sync. That degree is also measurable and seems constant. Rob is very helpful. Email him :-) http://pages.sbcglobal.net/rcrock/index.html Also, I made an interlaced video that has numbered fields that you can use to check the sync charateristics of your cams with an interlaced TV, down to 1/1,000ths of a second. When deinterlacing two cams, use the same advanced methods that the 2D videographers are now using, and render to full resolution pairs, then do all other 3D conversions from that. There is a huge growing list of advanced deinterlacer filters and plugins for VirtualDub and AviSynth, now, and the deinterlacer in Adobe After Effects also works very well. The best ones all seem to use a "motion detection" algorithm that theoretically leaves static areas alone, although I still see some blurring in static areas and use an "unsharp mask" filter to compensate with sharpening. Work is also being done with synced zoom. I think the LANC controllers are working for this, now, and I can get good simple zooms with just a standard remote pointed at both cams, but if a simple zoom is bad, it can be corrected in software. Brightness, saturation and other optic problems are common, but good software corrects most of the problems fairly easily. Good 3D video can be made from dual DV cams, but it does take some knowledge, good software, and work, and the process is improving almost daily. PKK | ||
| M.H. Rating: N/A |
PKK - try to record a fontane with drops of water falling down by any system witch you describe and do any software correction witch you want ... The result will be blured nothing in comparison to other systems. Or do you want to tell me that you will be able to see each individual drop of water in correct stereoscopic position for both eye ? It needs progresive recording and totaly HW synced recording. | ||
| Puppet Kite Kid Rating: N/A |
Welcome to the dilemma of progressive filming ;-) Deinterlacing debates are never ending. Just keep in mind that we can now manually sync two DV cams (without genlock) to better than 1/2,000th of a second for as long as they will stay in sync to that degree. Something that is interesting to think about, though IMHO, is motion smoothness in movies. I have alway heard that the human eye can see up to about 70 unique images per second. If those are progressive movie images, though, they will not have an inherent motion blur like you see in films or in _nature_. Only the objects that your eyes can focus on will not have motion blur. Everything else will. The only way you make a smooth motion with a movie is to blend those progressive images together and create motion blur. IOW, progressive images, even when played at 70 per second, will _not_ appear natural. They need to somehow be blended together to look natural. A still image, OTOH, is a different situation. Some people might prefer to see a water droplet "stopped" in mid air, and others might want to use a slower shutter speed and create motion blur in that on image. We have all seen examples of both. BTW, with 60i video, you have 60 unique images per second, and when those are separated, they are progressive, but they are half height. What you are really missing, compared to a 60fps progressive recording is resolution. The secret in good deinterlacing is finding a deinterlacer that leaves the static imagery intact, and "blurs" only the motion of each two adjacent unique images (separated fields). PKK | ||
| John Billingham Rating: N/A |
I can't resist: OK, you two, it's a DUEL ! I suggest you both post a SHORT clip of your best "Fountain" in the 3DTV group files, Then "Let the people judge"!!!!!!!!!! May the Best "StereoVideographer" win! John Billingham | ||
| Puppet Kite Kid Rating: N/A |
But, that's not fair unless we use cameras that cost less than $600 for both... or let's say $800, since I got mine *cheap* ;-) PKK | ||
| M.H. Rating: N/A |
To make a duel is a nice idea ... But it will be not fair as PPK says, becouse I will be not able to go with the total price of equipment used down to $5000. But I hope that Lightspeed people for witch I work will place some of our fountains test shootings (witch we use sometimes as test for genoclocking quality measurement) on the WWW for download. I am afraid that judging the winner will be realy a very subjective task. My personal experiences indicate that sharp genclocked progresice image beet anything. There is no place for any blure in stereoscopic movies becouse the eye is trying to focuse on the blured objects in the stereo space. | ||
| Puppet Kite Kid Rating: N/A |
Of course Michal is correct. If you have the money, get genlocked cameras. They are much better, and especially if you are working professionally, it's the only practical solution. People like myself, who are playing around with cheap consumer cameras, are basically experimenting... or having fun. The results usually take more time and work than what is practical. It is only a good solution for people with lots of time, but not much money ;-) The motion blur issue though, is where I seem to disagree. I will be watching Nascar3D at the Imax theater, tomorrow. Like many 3D movies, it was probably filmed at 24p (for each eye). In that film, you will see "motion blur". You will either be focusing on a moving subject that the camera is "following", which will _not_ show motion blur, but the background will, or you will be focusing on the static parts of the scene, which will _not_ show any motion blur, but the fast moving subjects will. Shooting in 60i, like I do with my cheap cams, provides you with 60 unique images per second, per cam. The fields have perfect integrity, and can be separated to 60 progressive images per second (after "de-bobbing"). Using the best deinterlacers, the subjects in motion can then be "blurred together" back to 30p, but the static subjects "left alone". You end up with something very similar to 30p filming with a slow shutter speed or "30p film" if there was such a thing, although there is still a bit of artifacting in the 60i conversion method(s), but it really isn't too bad, IMHO. Again, shooting progressive imagery is a very debated subject, and people are constantly trying to get a "film look", because the imagery tends to be too "stark" if you use fast shutter speeds. The common desire is to create a "motion blur" effect like you will probably see in the Imax Nascar 3D movie :-) PKK | ||
| Puppet Kite Kid Rating: N/A |
I just found a fairly good article on the new "Nascar 3D" movie. It has a good paragraph specifically including aspects of "motion blur", and I have pasted it, here. The entire article is actually quite good, and is found here: http://www.theasc.com/magazine/mar04/sub/index.html From that article: Strobing is another consideration in the Imax format because the negative's path is horizontal. "Objects moving across the screen too fast will be almost unwatchable," says Neihouse. "In order to reduce the amount of strobing, you must either slow the subject down - not an option on a NASCAR film! - or overcrank the camera to slow the subject down on the screen. But Simon wanted to show the fantastic speeds these cars a capable of, so we decided not to overcrank anything. By panning with the cars and keeping them in the same relative position in the frame, we were able to reduce [strobing], and by following the cars we were able to make the background blur out, which enhances the feeling of speed. At other times, we simply let the cars drive through frame, allowing the motion blur to add to the sense of speed. Another trick we used was to film the cars either coming at or going away from the camera. We never had to undercrank the camera to make the cars seem faster. They were plenty fast at normal [filming] speed!" PKK | ||
| Ged Rating: N/A |
I've just started shooting 3D video with twinned Sony PC105s and Rob Crockett's LANC Shepherd. Look at the following webpage for details: http://web.onetel.net.uk/~gedburnell/PC105.htm I use StereoMovie Maker (Fantastic freeware program) http://www3.zero.ad.jp/esuto/stvmkr/indexe.htm To combine the two streams into side by side format and view on the PC, Or interlaced for TV viewing. Michal, The only trouble with genlockable video cameras is the size of them! IMO for most shots they are an unneccessary expense. I can sync my non-genlocked camcorders to within 10 microseconds. Put another way a car travelling at 70mph (110KM/H) would move .3mm between shots! I'm now off to go find a fountain to shoot with my twin PC105s using a high shutter speed and mis-sync of 10 microseconds. I'll upload it to the Yahoo 3DTV group when done. Ged | ||
| M.H. Rating: N/A |
Ged: But you will shot interlaced anyway ... In addition the DV compression mixed partialy information from left and right interlaced lines during YV12 4:2:0 qanatization ... Even at perfect sync. you fall in trouble without progressive scan. On the other hand a perfect system with absolutely genlocked images with progressive NTSC per eye resolution does not need to be very big and does not need to consist from 2 genlocked cameras :-). | ||
| Ged Rating: N/A |
I think the Interlace vs Progressive scan is an other issue and is as valid in 2D video. But it is obvious you are talking about a professional setup. Whereas I am talking consumer level equipment. All I'm saying is excellent results can be had at a fraction of the cost of pro equipment. Ged | ||
| Ged Rating: N/A |
I have uploaded a clip which was filmed with a recorded mis-sync of 1/99999, 10 microseconds (possibly even better but this is the limit of the LANC Shepherd's display.) The shutter speed was 1/250, 50i fields/second but bob-deinterlaced to 50 progressive FPS. I'll let others be the judge of sync accuracy. http://web.onetel.net.uk/~gedburnell/PC105.htm Ged | ||
| Puppet Kite Kid Rating: N/A |
So, you are always at approximately 1/2000th to 1/2,500th of a second sync accuracy at power up, so you had to wait about 10 minutes after powering up to get that degree of sync? Even 5 minutes is usually enough for me to shoot just about anything I do, aside from some time-lapse (can't have everything for under $1,000, now can we ;-). Also, 1/2000th of a second accuracy is probably even plenty good for most subjects, especially if you do a field separation and frame blend in post. Obviously, my next cams will have LANC. I will probably buy them after I wear out one of my cams from turning it off and on too many times, getting them into close manual sync ;-) PKK | ||
| John Billingham Rating: N/A |
For God's sake, PKK, just wire the damned things together. (a few cents of wire vs. a replacement camera = "no brainer"!) THINK what it is that Rob's wonderful box ACTUALLY does???????? Best Wishes, John Billingham PS. GED, nice "Tap 'n Bowl" but it doesn't count as a "fountain"! | ||
| Ged Rating: N/A |
JB, I'm sorry I don't qualify for the test, my original idea was to be a shot of me in the bath creating my own fountain. But I thought this might be too shocking even for a NYer. BTW I've found that 1/2000 sync is more than adaquate for virtually all subjects. My Aquarium video was shot with a mis-sync of approx 1/2500. I can't tell any difference between channels. Ged | ||
| Puppet Kite Kid Rating: N/A |
I think any subject (or camera pan) that is moving so fast that a 1/2000th of a second mis-sync could potentially show a 3D motion discrepancy would produce inherent motion blur, so it won't matter anyway :-) Computer generated motion blur from frame blending always makes mis-sync less noticeable, so I am using that fact as basis for expecting the same with inherent motion blur. If you have motion blur, you can *get away* with some very large mis-syncs, even up to 1/120th of a second (the worst possible with 60i cams if you are willing to field sync in post). You should always see motion blur if your subject (or camera in a pan) is moving fast enough and your shutter speed is slow enough. Obviously, this is many times a challenging "mix". Even the professionals are always struggling with it with their expensive cameras. Fast moving subjects will "strobe" if they do not have motion blur, and will _not_ appear "natural" in a movie. It's "catch 22" and also is why video or films never produce good still pictures. Virtually all films (motion pictures) and video with *proper* shutter speeds have inherent motion blur. Obviously, starting off with perfect sync (or nearly perfect) is a good idea. Field syncing is never needed. Also, PAL camcorders are bound to be a bit more challenging (50i vs 60i). 1/120th of a second is the worst possible mis-sync with NTSC cams and 1/100th of a second of a mis-sync is the worst with PAL. That could possibly show an irritating 3D motion discrepancy, even with frame blending. PKK | ||
| John Billingham Rating: N/A |
No, Ged, We do it here too, (but , of course, we have the sense to "direct" the "fountain" OUT of the bath!) WE aim to please. YOU aim TOO , please! There is a rumor you are selling 3D "Potato Men" DVD's, Best Silly Wishes, John Billingham |
