3D on the Cheap, Part 1
Funny how a whole world can exist, and you may know nothing about it. So it is with 3D video on YouTube, a subject that I knew nothing about until a Fujifilm FinePix Real 3D W1 camera and a set of red/cyan anaglyph glasses fell into my lap. I said, "Cool," and then a second later, "What the heck can I do with these?" Turns out I can shoot fun 3D videos of my daughters, upload them to YouTube, and try to convince friends and family to spend a buck or two for their own pair of red/cyan anaglyph glasses.
Having said all this, I recognize that the prospect of uploading family videos to YouTube might not convince you to invest the 10 minutes or so you'll need to read through this article. Well, between the fluff pictures of my daughter, and screens of some some software you probably haven't heard of before, I'll try to filter in enough stealth learning about how to produce 3D videos to make it worth your while.
| Related Links |
||||||||
|
Specifically, in this article, I'll discuss how to shoot, edit, and upload 3D videos to YouTube, and what you'll need to view them from that site. In two weeks, I'll look at how to produce for Nvidia's 3D Vision system and review that system, which uses active shutter technology to deliver the type of quality available in movie theaters and which you'll see in the 3D TV systems just starting to appear in our living rooms.
Let me chum the waters with the one of the best 3D videos that I have seen on YouTube. You'll need red/cyan anaglyph glasses to view the video, unfortunately, but if you can dig one out of your kids' cereal boxes, it will be worth the effort.
My focus for this article is the editing workflow, and uploading to YouTube, but I'd be remiss if I didn't at least touch on the acquisition aspect. As mentioned, I used the Fujifilm FinePix Real 3D W1 camera for my 3D video shots. As you can see in Figure 1, the camera features two lenses, simplifying the shooting and capture chore. The downside was that the maximum capture resolution was 640x480, which is OK for YouTube and even DVD output, but higher-definition videos are sharper, which delivers better quality 3D, even on YouTube.
Figure 2 shows a DIY rig from www.3dphoto.net, a very good resource for information on 3D photography and video acquisition, using two Aiptek A-HD 720p video cameras. You can find instructions for producing the rig here. Popular Mechanics details how to produce an inexpensive 3D rig with primitive steadycam features here.
Figure 3. Ikan's stereoscopic 3D camera rig, demonstrated by Tim Dashwood at ProFusion (figure from ikan's blog).
Camera support vendors are starting to enter the picture. For example, at NAB, ikan showed a stereoscopic 3D camera rig that mounted two camcorders side by side, though the product isn't yet listed on ikan's website. I would expect more of these to appear in the near term from ikan competitors.
Of course, if you have $21,000 to spend, you can buy Panasonic's twin-lens Full HD 3D AG-3DA1 camcorder. You can also read about some other lower-cost 3D camcorders here.
As mentioned, my goal here isn't to be comprehensive. I just want to make the point that if you want to shoot in 3D for distribution via YouTube, you have options. I also want to make two other points.
First, creating a 3D rig or even using a camcorder like Panasonic's 3D camcorder is fraught with technical issues such as how far apart your cameras should be, and whether their focus should converge at some point in the distance or be setup in parallel. For an overview of these issues, check out Tim Dashwood's A Beginner's Guide to Shooting Stereoscopic 3D.
If you just want to dash off some 3D videos for YouTube, you can skim the article, build your rig, and start shooting. On the other hand, if you're trying to create serious work for clients, you need to understand and incorporate these concepts into your setup and your shoots.
Second, the creative aspects of shooting 3D are equally important. For example, if your video doesn't contain any visual cues about depth, you'll get no 3D effect even with a perfect rig. If your primary subject touches the borders of the video, you lose the 3D effect temporarily. If you shoot video with lots of red content, it will totally desaturate in anaglyph displays. On the editing side, you have to use longer sequences with fewer cuts, since it takes a few seconds for the brain to create the 3D image from the left and right inputs. Color correction can create problems with video bound for anaglyph display.
These are just the easy ones. Google "shooting for 3D video," and you'll see a host of articles, tutorials, and other content on the subject. You can also scan through YouTube and see which 3D videos work well and which don't. Just understand that if you shoot for 3D the way you shoot for 2D, the results may not be optimal.
OK, 'nuff said on shooting, now onto the editing. Here I used three programs: Roxio VideoLab 3D, which is a consumer-focused OEM solution showing features that are expected to be available as part of a consumer solution shipping this fall; Stereo Movie Maker, which is a free program that you can download; and Adobe Premiere Pro CS5, though you can use the same technique with any video editor.
The Fujifilm FinePix Real 3D W1 camcorder stored the left and right streams in a single AVI file. Working with the Roxio VideoLab 3D project was simple: I just input the video, and the program asked me to identify the source material, which I did.
From there, I edited my project as normal in VideoLab, and then used the program's Upload to YouTube feature to share my work. Obviously, there are other outputs availableincluding DVD, Facebook or a digital file in a number of digital formatswith the presets shown in Figure 6. This makes it easy to format your video for your 3D video device, which might require anaglyph, side-by-side or top/bottom images. In contrast, when working with Premiere Pro or other editors that aren't 3D aware, you have to customize your sequences so that they match the input expected by your 3D viewing system, in this case YouTube.
Step one in Premiere Pro was to figure out the format that YouTube would accept. Fortunately, YouTube makes this simple with a technical note titled "Getting Started: 3D Content." In terms of layout, YouTube expects a side-by-side display, with the right camera image on the left and the left camera image on the right, as you can see in Figure 7.
If you peruse the document, you'll see one other sequence-related requirement: "For best results, the combined image from the two videos should be 16:9 resolution (distorting your frames as needed. The aspect can be corrected by the player)." More on this in a moment.
Figure 8. StereoMovie Maker was quite competent, but I just needed it to create separate streams from my FinePix file.
When working with two different cameras, you acquire the files separately, import them into your editor and synchronize them on the timeline. However, though the AVI file captured by the FinePix contained two video streams, Premiere Pro saw only one of them. So I had to use StereoMovie Maker to separate the two streams.
As mentioned, StereoMovie Maker is a free program, which you download and run by clicking the EXE file. This means that there's no installation, which I like. If you don't have a Motion JPEG codec installed on your computer, you won't be able to play the FinePix files when you load them into the program, but StereoMovie Maker will direct you to a site where you can download demo versions of several Motion JPEG codecs.
For a free tool, StereoMovie Maker is actually quite competent, with the useful ability to synchronize and adjust the left and right streams captured by separate cameras, as you can see in Figure 8. However, all I needed it to do was create separate left and right streams, so I dragged the video file from the FinePix in and chose File > Save Left/Right Movies. The next prompt is for compression format, and I used the Motion JPEG codec that I had just downloaded and installed. StereoMovie Maker created the separate AVI files in just a few moments.
These files loaded fine into Premiere Pro, so it was time to create my sequences and start editing. Mindful of the Google request that the combined video should be 16:9 resolution, even if I distorted my frames, I built one sequence at 1280x720, and distorted the videos using the Motion controls until they both fit side by side in the 720p frame as shown in Figure 9.
I also created a side-by-side sequence without distortion that turned out to be 1256x480, since StereoMovie Maker output two 628x480 files. I rendered both sequences into high quality H.264 format at the same resolution as the sequence, and uploaded them to YouTube, along with the video captured by the left camera to show what a 2D rendition would look like. Note that when you upload a 3D file, you have to include the string "yt3d:enable=true" in the Tags field, as shown in Figure 10.
I played the two 3D files and saw no significant difference from a 3D perspective. The 720p file looks clearer, probably because YouTube produced a720p HD version of the file, where the lower 1256x480 input maxed out at 480p. You can check out the three files on YouTube and draw your own conclusions.
As you can see in Figure 11, you have your choice of multiple 3D viewing formats, selectable via the list box on the bottom right. So get out your 3D glasses and have a look. That's Whatley, by the way, my eldest, just back from two weeks of ballet training at the Rock School in Philadelphia. Please excuse her camera shyness.
| Related Links |
||
|
The discerning viewer will note that I broke several of the 3D-related shooting rules that I stated above, including dressing Whatley in a reddish shirt that turned yellow in red/cyan anaglyph format, and by having the bulk of the 3D motion enter and exit the frame quickly. For this reason, Whatley's arms and legs never seem to hover outside the monitor the way the knife-wielding youth does in the first YouTube video we viewed, though the furniture in the room looks nicely set off from the background.
Come back in two weeks for a discussion of Nvidia's 3D Vision and better videos.
















