Bringing digital video editing inhouse

The inhouse creation of video used to be the private realm of specialized studios. Now, video creation is following the growth patterns of desktop publishing and desktop graphics. Studios used to do all the video editing with tape decks and splicing, but now the process is digital. When video editing first became a digital process, the Macintosh was the platform of choice. Thus, the Mac was the foundation for all the development of powerful desktop video editing software.

The Mac OS continues to dominate the market, but advances in operating systems for Intel-based PCs are turning the tide. Windows NT 4.0 is spurring the PC's acceptance in the video editing market as more people discover that NT is stable, powerful, and can grow as demand increases. The true multitasking and symmetric multiprocessing (SMP) functions available in NT also allow significant performance gains.

Motion picture and TV studio systems might stick with the Mac because of inertia. However, as corporate systems perform more video editing, these systems will probably be NT based. With NT 5.0, the addition of multiple monitor capabilities and extended driver support will strengthen NT's position.

Linear and Nonlinear Editing
The two kinds of video editing are linear and nonlinear. With linear editing, you record each piece of video and audio in sequential order. A crude example of linear editing is to connect two VCRs and tape from one to the other. You play a segment of your original tape on one VCR and record it on the other. When you get to the end of the segment, you stop recording and change the tape in the source VCR. You then play and record a segment of the next tape that is in the source VCR.

Nonlinear editing is the more popular type of editing because it is flexible. You don't need to worry about the order in which you record the video and audio because you can manipulate each segment independently and move it along a time line. After you have massaged all the segments, generated the titles and effects, and perfected the sequences, you can record the final video either digitally or onto a tape.

You prepare video with a five-step process.

  1. Capture the subject via videotape or film.
  2. Transfer the raw image to hard disks for editing.
  3. Edit the material.
  4. Add graphics and effects.
  5. Format for a medium (i.e., TV, CD-ROM, or Internet video).

You also need to consider cost. The video industry has a saying about video production: You can get it fast, you can get high quality, or you can get it cheap--but you can have only two out of three. I have seen staggering price quotes for productions, and I now understand why movies are so expensive to produce.

However, I wanted to create a way to turn out videos quickly and efficiently, get great quality, and save money. To meet these goals, I had to bring as much video production inhouse as I could afford.

One of my projects is to produce a videotape demonstration of Windows NT Magazine's Windows NT Solutions Directory (http://www.winntsolutions.com). Because the video will be used on the Internet and in other ways, I needed top-quality and high-performance editing and production at a reasonable cost.

First, I had to decide how to edit the video. The standard method is to use a Macintosh running Avid Technology software. Avid Technology controls about 95 percent of the digital video editing systems in the studio environment, and nearly all these systems are Macintoshes.

But I'm an NT guy, so I needed a way to do this project with NT. I found several choices in the rapidly changing landscape of products. Because the NT digital video editing market is new, not all devices are compatible, nor can software compatibility be guaranteed. New digital video products enter the market every day, and each company has its own approach to compatibility. With this dynamic market in mind, I knew that good technical support, demonstrated performance, and reliability were critical.

Editing Software
I found about a dozen video editing packages for Windows 95 or NT. On the low end of the video editing software spectrum, inexpensive packages provide basic editing capabilities but also have limitations that drive a professional editor crazy. An example is Corel Lumiere, which cannot import files larger than 500MB. This file size seems large, but it is only about 45 seconds of uncompressed video. You can import as many 500MB-or-less files as you want, but the size limitation is a hassle.

One large step up, Adobe Premiere, from Adobe Systems, has the largest market share, with equivalent products on NT and the Mac. This product is easy to use and flexible, and has more plugins than any other product. Adobe Premiere is fast and works with more capture cards and operating systems than any other program I found.

But the product's high-end capabilities are limited. For example, you can't render video in real time. As a result, the system takes extra time to assemble the video and audio layers to create a single preview. Depending on the complexity of each layer, the number of layers, and the length of the video, a preview might take several minutes or even hours to complete. If you are creating complex video projects, this wait is counterproductive. However, this limitation is where NT's scalability comes in. Adding more hardware power dramatically improves rendering times.

Another limitation of Adobe Premiere is that the quality of output decreases when you must resize clips inside the program for final output. With this software, you will always want to set the original clips at the same size as the final output. Both Speed Razor and MCXpress adjusted the sizes flawlessly.

In the upper range of the product spectrum is in:sync's Speed Razor. It has more power, more flexibility, and more hardware demands. At the high end is Avid Technology's MCXpress for Windows NT, which oozes raw power with few limitations. These professional products offer the most power, but you must master these products' learning curve.

The basic differences between the low-end and the high-end products are the import and export features and the number of simultaneous tracks available. Speed Razor really shines in these areas. With its unlimited video and audio tracks, you can make your mix as complex as you want. Speed Razor also lets you drop any type of graphic onto its dedicated drive and will configure the graphic on the fly to let you use it with the editor. This feature was particularly nice with .gifs, .jpgs, .bmps, and .eps files. I didn't need to worry about importing these files--I just dropped them onto the dedicated drive.

MCXpress doesn't have as many automatic options in the import and export process, but it has outstanding titling and special effects. Speed Razor offers additional special effects and titling, but only as options. Adobe Premiere has a wealth of add-ins because of its popularity with users. It uses more types of inputs and capture cards and has direct plugins for export to RealVideo and other plugins for direct work with other graphics programs. Unfortunately, you'll probably need some of the plugins to get all the functionality you want.

The Hardware
Digital video editing requires a powerful system. To store uncompressed video, you need 11MB of hard disk space per second of video, and audio requires additional space. Large, fast hard disks are the order of the day. Audio/visual (A/V) hard disks have large buffers and high spindle speeds to store the data without dropping any frames. The minimum hard disk size is 9GB, and even with this size, you will regularly send clips to a high capacity removable drive, such as MO (magneto-optic), SyJet, or Jaz.

Video is captured at 30 frames (NTSC) or 25 frames (PAL) per second. Capturing video requires extreme processor performance, or frames can be dropped. I tried a 200MHz Pentium Pro, and then added a second processor. In this case, because of the built-in MMX technology, a dual 266MHz Pentium II would have been even better. Using NT has its benefits--I can throw a lot more hardware at the challenge and make it work well.

Each video editing software package I tried claims to work with 32MB of RAM. My system came with 32MB of RAM, which I quickly bumped to 96MB. The additional RAM made a huge difference in the system's performance.

After you configure the hard disks, the processor, and the RAM, your next challenge is to configure the capture card. Video devices, such as cameras and VCRs (called decks in the video industry), usually have one of three output types (I list these in order of picture quality, starting with the lowest quality): composite (using RCA jacks), separate video (S-video), and component (using multiple BNC connectors). The low-end cards, such as the Intel Smart Video Recorder III, have both composite and S-video inputs. The PVR-2500 Perception Video Recorder card from Digital Processing Systems (DPS) and the Targa 2000 RTX board from Truevision have all three types of input jacks. Other capture cards are available, but these three cards are the most common in their respective markets.

Prepare to allocate a lot of physical space for your capture card. For the Perception card, with its main capture card plus its daughtercard, you will need two PCI slots. I had five PCI slots, which are now occupied: one for my video card, one for my SCSI card, two for the Perception card, and one for my sound card. With these cards installed, I don't have room for my network card.

The next critical component you need is the sound card. You can use any sound card with Intel's Smart Video Recorder, which worked well with my standard 16-bit sound card. Obviously, the better the card, the better the sound will be. To use Speed Razor, you need to add DPS' A4V, Antex Electronics' 23e or Studio Card, or at a minimum, Turtle Beach's Tahiti or Monterey sound cards. Truevision incorporates audio into its Targa 2000 video boards. The difference among the sound cards is in matching the audio to the video to avoid a poor lip-synching job. The high-end boards use the industry standard SIMPTE time codes to make matching audio and video easier.

Exporting Computer Video
Because so much of my work is based on computer training, I need to transfer my computer video to videotape. You can export video with a scan converter, a video card with analog video out, or a screen capture. Higher-end video cards with analog video out perform comparably to scan converters, which cost more than $20,000.

A scan converter takes the analog signal that usually goes to the monitor and changes it to NTSC video--you can actually watch your computer's screen images on a regular TV. Several devices claim to perform this function. Scan converters usually work with any operating system because they are external devices that simply change the type of signal.

Video cards with analog video out bring the scan converter into the computer system with a set of chips on the video card, plus special connections. Although several outstanding cards are on the market, I couldn't find any that have NT drivers to take advantage of the video-out capabilities.

Screen-capture software takes pictures of the screen at a rate of 3 frames per second to 30 frames per second and saves these images on your hard disk as .avi files. This process does not require analog-to-digital conversion, and the quality is very high.

Lotus ScreenCam (screen-capture software) outperforms both analog video out and scan converters, but requires offline preparation of the video. It also has some compatibility problems with NT. I chose the ATI Technologies 3D Xpression+ PC2TV accelerator for live capture and HyperCam (another screen-capture program) from Hyperionics for offline capturing of video screens.

To export the video, I recommend at least a 166MHz Pentium if you choose Lotus ScreenCam or HyperCam. If you use anything less than a 166MHz Pentium, you will encounter quality problems. For example, when I ran HyperCam on my 100MHz Pentium, I could see only half the letters on each line of the captured screens.

Making It All Work Together
Now, for the challenge: solving the compatibility issues. Not all editing software works with all cards on all platforms. The Smart Video Recorder works beautifully in Win95, but as of this writing, does not have NT drivers. The Perception board works and integrates well with Speed Razor, which will run on Intel or Alpha. But Perception won't work with MCXpress; MCXpress works only with Targa boards, which have several levels and prices. Also, MCXpress lacks an Alpha version. Adobe Premiere is independent of the type of capture card, which contributes to the card's tremendous user acceptance.

Is a Standardized Format Desirable?
Nearly everyone who uses computers complains about compatibility: "I wish word processor A used the same document format as word processor B." Or, "Why can't I import that information into my accounting program?"

Video file formats make other file incompatibility problems pale in comparison. Among the .avi formats, you'll find so many differences in the files that you spend a lot of time just trying to make things work together. Maybe these compatibility problems are the reason the video editing industry still sells turnkey systems, which is typical in a market that hasn't matured. Until all the manufacturers can get together, set standards, and apply them, cross-platform and cross-application compatibility is only a dream. Many companies, such as Avid Technologies and in:sync, create outstanding products, but see themselves as the only players in the market. They charge outrageous prices for their hardware and software, have inadequate support, and have proprietary formats that are difficult to work with. This situation reminds me of AutoCAD.

The video editing market is ripe for a revolution. Software such as Adobe Premiere is sufficient; but for hardware, I'm waiting for a company such as HP to step in and standardize the market and the drivers, and consolidate the output files as HP did with its laser printers. Maybe Microsoft's DirectShow and DirectDraw APIs and Active Streaming Format will force manufacturers to standardize output.

The MCXpress and Targa solution turns your system into a turnkey editing system. These products are so closely integrated, they seem to have been made for each other. Speed Razor gives you more video card options, because it uses either the Targa boards or the Perception board. It will also use different sound cards.

I set up three systems: one with MCXpress and a Targa 2000 RTX; another with Speed Razor, the Perception boards, and an A4V sound card; and a third with Adobe Premiere, working only with previously captured video and audio clips. Now the fun was beginning.

Setting Up Windows NT
Most video editing systems use dedicated hard disks for storing video clips. With the Perception card from DPS, the hard disk is connected directly to the capture card. You don't format the disk with NT, but through the Perception software. Although the disk is not NT formatted, you can see it and use it in Explorer and other applications. The drive is labeled P:\, and you can copy and move files back and forth from your system drive to the dedicated drive. This capability makes the drive very simple to work with and provides the necessary performance to capture video. Although NT can use the drive, Disk Administrator cannot manage it. The Targa system and Adobe Premiere do not require a special connection for their data drives, but the companies recommend fast drives dedicated to the capture and storage of video files. None of the software I tested let the system volume also act as a data drive.

The drivers for the audio and video capture cards have been a problem. Recently released drivers allow the use of high-end audio cards as system sound devices, but when I use them, I can't play my music CD-ROMs on my system. The Intel Smart Video Recorder III lacks an NT driver, as does the DV Master from FAST Multimedia. Without NT drivers, the capture process has to run through a Win95 system, and I have to endure all the attendant problems.

The only thing I needed to change about my NT system software during my tests was the amount of virtual memory. Even with 96MB of RAM, I had problems running the different programs until I changed the size of my paging file. The NT default is RAM plus 11MB. For optimum application performance, virtual memory is usually set 1.5 times to 2 times greater than the amount of RAM. Surprisingly, for capture, DPS recommends a paging file smaller than the amount of RAM--8MB less than the physical RAM. Capturing data through a capture card requires the ability to write at least 4.5MBps to 5MBps. Using a paging file during capture makes matching that rate difficult. I made the requested change, and the capture functions worked very well.

However, the paging file was insufficient when I was using the editing programs. When I put a 500MB paging file on my data drive, everything smoothed out, the performance increased, and reliability was no longer a problem.

I set the paging file to 88MB while I was capturing video, and added a 500MB paging file on another drive while I was editing. This setup works well, but this configuration is a hassle.

DV Format vs. Motion JPEG
Most of the capture cards that feed analog or digital video into the PC use Motion JPEG, which combines many .jpg files, one for each frame. You then edit these files and process them for output. The system then changes the format to .avi or QuickTime (.mov). This change makes editing a breeze, because you can deal with each frame independently of the others. Nonlinear editing uses this method. In fact, most nonlinear editing systems must translate any other format to Motion JPEG before editing can occur. The biggest challenge associated with hardware that turns video into Motion JPEG is that the same hardware cannot capture video for live broadcasting.

One system can accept a pure Digital Video (DV) signal and use it, without modification, for nonlinear editing. DV Master from FAST Multimedia lets you capture, transfer, and edit in DV format. Because DV is the product's native format, you can speed up the process, maybe even to the point of going live from a DV camera into the PC. I couldn't test this capability because it wasn't yet released for NT.

The Shoot
I finished the Windows NT Solutions Directory, and I scheduled a video shoot. Eight takes later, I completed the shooting and had the pieces I needed to put together a dynamite clip.

Editing
In every editing software product I worked with, only the clip on the highest numbered track is visible. In other words, you will see what is on track 1, as long as nothing is on track 2. If track 2 is nontransparent video, you will not see what is on track 1. If you want a title on the top of a video clip, you put the video clip on track 1 and the title on a transparent background on track 2. If you want to switch to a different clip, such as a picture of the computer screen, you can simply lay the screen shot on a higher track number. You don't need to delete the section of the first clip that was behind the new clip.

In MCXpress, with its limitations of two video tracks and up to four audio tracks, you set the video clips with precise edit points end-to-end on one track, and then put the effects and titles on the other track. With unlimited tracks for both audio and video in Speed Razor, you have more flexibility in how you layer your video and audio tracks, which is helpful. In Adobe Premiere, you get a compromise, with two standard video tracks, a transition track, and unlimited superimpose tracks (S tracks). The S tracks are for creating multiple layers with transparent backgrounds and titles.

I took the clips and captured them with the DPS Perception card and then the Targa board. Each card comes with proprietary software just to facilitate the capturing process. After I made the correct settings for the audio and video capturing, the editing was easier than using a VCR.

Next, I arranged and trimmed the clips, and added my titles and graphics. I added my captured screens from HyperCam, laid them over the video track, and left the audio track untouched.

Rendering and Exporting to AVI
The process of rendering takes a set of audio clips, video clips, graphics, and effects, and puts it all into one video that you can watch from one end to the other. You can click any place on the time line and see what your video will look like at this point. This function required more time than nearly any other process. This area is where NT and serious horsepower really kick in. Put a dual Pentium Pro or Dual Pentium II on the task, and you'll cut your time in half.

Usually, when you are going to export the final project to an .avi file, the system renders the video again, sets all the clips to the same size, sets all the audio to the same values (mono, stereo, KHz, etc.), compresses the whole video, and then writes it to your hard disk. Depending on the compression routines, this process can take hours. Obviously, if you are going to use videotape as your final product, you can skip the .avi file and just dump the project to tape.

Recommendations
No matter which way you decide to go, I have some recommendations for making video editing much easier. First, get heavy-duty hardware if you are going to do serious work. Don't use anything less than a 200MHz Pentium Pro with 64MB of RAM and at least five PCI slots, plus two or more ISA/EISA slots that the PCI boards will not cover up. Hard disk space is also a real necessity. A 9GB A/V drive is the minimum I recommend, plus your 2GB to 4GB system drive. Make sure your system has plenty of drive bays for internal drives, because they are far less expensive.

All three video editors performed well, once I got them fully configured. An advantage to the Avid MCXpress is that an editor who has used a Mac version will have almost no learning curve. However, even editors with extensive experience will have a minor learning curve with Speed Razor.

Keep in mind that if you are starting from scratch, all of these systems require a serious time commitment. Video editing is not word processing. You must practice to acquire the necessary technical skills and to master the programs' intricacies.

Which product will I use next time? The answer depends on the project. You can't go wrong with any of them. I like Speed Razor's unlimited layers of audio and video, which make pulling video from different sources and adding multiple effects unbelievably easy. However, the titling and onscreen graphics creation in MCXpress outshines Speed Razor's. Even so, I might use Adobe Premiere, which is the easiest to set up and learn and has all the power I need.

Contact Info
Adobe Premier
Contact: Adobe Systems * 408-536-6000
Web: http://www.adobe.com
MCXpress for Windows NT
Contact: Avid Technology * 978-640-6789 or 800-949-2843
Web: http://www.avid.com
Speed Razor
Contact: in:sync * 301-656-1700 or 800-864-7272
Web: http://www.in-sync.com