The Real Deal

Stephen F. Nathans

February 2003 | Time was when fast DVD turnaround meant days at the replicator.But these are simpler days,with simplified tools enabling simplified tasks: capture, encode, and burn. Cheap recorders, fast PCs, and consumer DVD tools promising "real-time" authoring are rapidly turbo-charging desktop DVD creation. But how real is real-time, and how readable are real-time-written discs?

Most people who decide to be writers of one sort or another can point to books that inspired them to pursue that course. For me, two that I read within a year of one another (at ages 13 and 14) stand above the rest: John Irving's The World According to Garp, because it portrayed a novelist as something to be; and E.L. Doctorow's The Book of Daniel, because it awakened me to politics and history, as well as how much a novel could say, and how electrifyingly it could say it.

Of course, this kind of experience with books isn't confined to writers. How many future riverboat captains grew sea legs after reading Life on the Mississippi; how many environmental activists found their calling in Rachel Carson's Silent Spring; and how many rich kids discovered root and reason for their self-pity in The Catcher in the Rye?

But what about the books that capture our imaginations for more visceral reasons, that simply hook us on reading because they sweep us away so completely from the mundanities of the world around us? For me (and countless others, I'm sure), one book proved peerless in that regard: The Count of Monte Cristo. I can't wax quite so philosophical about The Count as I might about Garp or Daniel, but I doubt either of those books would have meant much to me¡Xor found their way to me at all¡Xif The Count hadn't found me first. Recently, after viewing the 2002 screen adaptation of The Count on DVD, I dug out my copy of the novel, and couldn't quite put my finger on what was so great about it. I know I originally read a different translation in a library copy (I bought my paperback years later mostly as a souvenir), but I doubt that's it. And I certainly found the DVD a powerful reminder of how completely the book held me in thrall¡Xhow it filled me with rage over the injustice done Edmond Dantes, and left me unable to do anything but read that book for a few magical weeks in the fall of '82.

So maybe it wasn't the aesthetics of the writing, which strikes me now as unexceptional, and it certainly wasn't an infatuation with post-Napoleonic France, or 19th- century Romanticism (no other book in that category ever made much of an impression). Which basically leaves the story¡Xand what a story it was! And how ironic that it should be the last thing that comes to mind.

That said, I'll take a great tale told by a less-than-great teller over the reverse any day. (Which is not to say Dumas' brisk, vivid prose is less-than-great¡Xit's just not the greatest thing about his book.) No amount of artful telling is going to make up for a trite or dull tale, but¡Xassuming the teller's shortcomings stay out of the way¡Xsome stories are so great they tell themselves. Take The Gospel According to St. Matthew¡X the last place you'd look for artful writing, but nonetheless a story for all time.

Most of the story we tell here concerns interactivity, Web connectivity, usability, feature-richness, and other elements of aesthetic and technical ingenuity that contribute to the success of a DVD-Video title or project. But ultimately, at the heart of the matter is a movie on a disc. Admitting a rather broad definition of "movie" (think of Macromedia Director's), maybe it's not a Hollywood movie, or even a commercial one. Maybe it's a home-made video, a training application, b-roll footage, or other video archived for later use¡Xbut in some way, it's video stored on disc to be retrieved later via a DVD-Video player. And if that let-the-story-tell-itself approach is what we have in mind, what's the simplest and fastest way to make it happen?

Simple: Capture, encode, and burn, and preferably, do it as quickly and automatically as possible. Fortunately, the makers of the growing class of "entry-level" DVD authoring tools¡XSonic, Ulead, Pinnacle, MedioStream, CyberLink, InterVideo, and SCM¡X have "direct-to-disc" DVD-Video creation in mind, too, and all of them would like to have you believe they've found the simplest and fastest way to make it happen. Of course, they have other ambitions for their tools, too, like making more accomplished DVD authors of their users¡Xavailing them of basic video editing and clip-trimming, original and template-based menu creation, encoding rate choices, and the like. But they also know that a key part of rudimentary DVD creation¡Xeven, arguably, for those DVD creators with more than rudimentary skills¡Xis fewest-steps-possible capture-to-burn DVD-Video recording.

Several of the vendors providing these tools claim "real time" for this process. A year ago, this notion would have seemed absurd. Streamlined as an authoring tool might be, it's simply not within the power of a piece of software to provide this kind of functionality. Reliable real-time capture has been a reality for some time¡Xeven with the cheapest PCI FireWire capture cards¡Xand obviously, any PC that can support a DVD Recorder can support at least 1X recording. But the real bottleneck comes in what some tools call the "rendering" stage, which mainly consists of encoding the video (captured as DV-quality AVI files) to MPEG-2. Even on a modestly fast Pentium IV, such as the 1.5gHz Dell we used as an in-house testing unit through the first 10 months of 2002, "rendering" speeds for highest-quality MPEG-2 (8Mbps) averaged six-plus minutes of rendering per minute of video. And that was a substantial improvement over software-encoding speeds achieved on sub-gHz PCs. But now, with the growing installed base of fast Pentium processors of 2gHz and beyond¡Xnot to mention the ultra-hyped, "hyperthreaded" 3.06gHz champion chips which offer virtual dual-processing¡Xit's not an unreasonable claim.

To test the reality of "real-time" DVD creation, we took two tacks. First, we surveyed the engineers and product managers responsible for these products about what makes real-time possible, taking into account both the hardware factors (capture cards, processors, bus speeds, hard drive speeds, recording speeds, and system resource allocation), and what can be done on the software side to accelerate the process. Second, we tested the direct-to-DVD functionality of seven entry-level Windows DVD tools¡XSonic MyDVD 4, SCM Dazzle DVD Complete, Pinnacle Expression, Ulead MovieFactory 2SE, InterVideo DVD Creator Plus, CyberLink Power- Producer, and MedioStream neoDVD (plus Apple's iDVD; see sidebar)¡Xto see how fast each tool could take a ten-minute video clip from capture to disc. In doing so, we considered both scientific and non-scientific factors: that is, not just how long each step (capture, render/encode, build image, and burn) took, but also keeping the clock running for how long (and how many steps) it took the program to get me through the process, given a pre-existing familiarity with the interface and procedure. If that isn't "real time," what is?

The race is on.

THE ROAD TO REAL TIME

So what is "real-time rendering," and what makes it possible at long last? To answer that question, we first have to look at how far we've come to reach the point where a novice user with $500 DV camcorder and a $1000 PC with a $29 FireWire card, $49 authoring tool, a $250 recorder (that may even come bundled with a $49 authoring tool), and $3 recordable disc can create a playable DVD-Video disc in something approximating "real time."

For the purposes of our discussion here, we can dispense with several elements of the equation quickly:

• PCs are cheaper and faster (more to come on that) than they used to be

• the ubiquity of cheap DV camcorders has made high-quality digital video recording available to a vast new class of users

• the IEEE 1394 (FireWire) interface and the cheap PCI cards that bring it to PCs avail just about anyone with a PC and sizeable hard drive of reliable digital video capture; real-time capture became possible with the introduction of the Pentium 3

• DVD authoring tools are vastly cheaper and simpler than they used to be, and many "consumer" tools are optimized for seamless "direct-to-disc" capture, author, and burn (more to come on that)

• DVD recorders are faster (1X eighteen months ago, 2X at this writing, 4X by publication) and cheaper (from $5000 to $250) than they
used to be

• DVD recording media is faster (mentioned earlier) and cheaper (from $17 in 2001 to $2-3 today for top brands) than ever

Time was when you had two viable choices for transcoding/rendering DV video to MPEG-2, the most resource-intensive work in DVD-Video creation, and the make-or-break step in any project. You could outsource or buy a hardware encoder. Both these approaches were costly, and required professional-level commitment and ambition, arguably, to be worthwhile, since effectively managing variable bit-rate encoding on a hardware MPEG-2 encoder isn't exactly a walk in the park. The advent of more powerful PC processors hastened the viability of software encoders¡Xwhich sapped the resources of weaker systems¡Xand made the DV-to-MPEG process cheaper and easier.

But it wasn't until Intel brought out 1gHz and faster processors that you could really do software encoding in reasonable amounts of time; for a year I used a 1.5gHz PC that took roughly six minutes per minute of video to render/ transcode DV video to highest-quality MPEG-2. And even with the onset of "consumer" DVD creation brought on by all these factors, high-quality encoding remained as important as ever, given the inevitable infusion of lower-quality (read: home movie) sources into the DVD mix.

GETTING REAL

Some software vendors would argue otherwise¡Xthat the real-time era arrived as soon as you could slip lower-quality MPEG-2 in under the real-time wire. And speedier encoding certainly has its place, even in the "pro" realm, where direct-to-disc tools should have a distinct workhorse appeal for getting digital video¡Xwhatever the quality¡Xonto DVD fast.

According to Stephane Desproges, director of applications engineering at neoDVD manufacturer MedioStream, that happened at 1.1gHz. Since he specs the "average" PC among the current (Q4 2002) installed base as a 1.1gHz Athlon AMD with PC133 bus, that means "the majority of systems sold today will support real-time DV to MPEG-2 encoding." He adds that these numbers apply to "100% resource availability" circumstances¡Xi.e., with "nothing else running."

Marc Williams, program manager for Dazzle, says his company's DVD Complete can achieve real-time DV-to-MPEG-2 conversion in "Fastest Performance" mode on a 1.4gHz Pentium 4, which still arguably qualifies most PCs that are likely to be used for DVD authoring today. In both cases, these estimates may reflect encoding bit-rates of 3-4Mbps, the lowest acceptable range for MPEG-2 or the use of fast, low-acuity encoders.

Across the board, software manufacturers agree that processor speed is the #1 issue in encoding speed, once you equal out other factors, such as the amount of motion in the video, since motion-intensive video taxes the encoder more. But there's little consensus on the exact processor speed at which rendering reaches real time. And since each software tool may define "quality" levels differently (8Mbps maximum, 3Mbps minimum), use different nomenclature, use different encoders, and set different defaults, that further complicates what users can expect using various tools. Most of the tools set the default encoding bit-rate in direct-recording mode at highest-quality 8Mbps, which may mean the hands-off user who doesn't tweak the settings will need a PC with significantly more power than the specs quoted previously. (Keep in mind different tools use different encoders, and a simple tool will use different encoders for different quality ratings, and may even use different encoders in the bundled and retail version of the same product.) For highest-quality 8Mbps encoding, the default in MyDVD 4, Sonic's Mark Ely says users will need a 1.6gHz for the software to encode video in real time.

According to Marc Williams, the key factors in encoding/rendering performance are "CPU speed, bus speed, and memory speed. To a lesser extent," he adds, hard disk speed and throughput factor into the equation. "Since encoding is CPU-intensive," Williams continues, "higher CPU rates will improve encode times. However, after a certain point, higher bus/memory speeds may make more of a difference. The behavior of a particular CPU and speed depends on the overall system configuration."

DV-quality AVI files (the source format for this entire discussion) consume 13.5GB of memory per hour of video, and even a one-hour DVD may be built from several hours of source footage. Consequently, the necessity of DV capture is the availability of an enormous amount of hard drive space for staging the video. Users may meet these needs in a number of ways. In my testing, I used the 90GB storage partition (D:) of my 120GB factory-installed hard drive. Where typical C: drives may come up short, users may add ATAPI, FireWire, USB 2, or (much less frequently today) SCSI hard drives for staging/storing their captured video. Subsequently, the software must then draw the video from that source at the encoding stage. Opinions vary on how the bus speed of the stored video source may affect the process.

"The faster the PC can access the DV frame, the faster we can process it," Ely says. "This is more of a limiting factor on a faster CPU where the CPU is no longer the limit, but the disk is."

"For encoding times, this is not as important as it used to be since the speed of hard drives has kept up with processor speeds," says Brian Lane, product manager for Pinnacle's Expression. "For example, a UDMA/66 ATAPI drive can easily keep up for the encode when using a 1gHz or less. When you start getting into 1.7gHz speeds, UDMA/100 drives are sufficient and UDMA/133 for 2gHz and above should be more than enough."

Williams comments, "As long as the minimum DV throughput of 3.8Mbps is maintained, for example, drive speed capability above that will not be used. However, there have been cases of FireWire hard drives interfering with a FireWire camera, which could affect results if those are used simultaneously."

MedioStream's Desproges adds, "When creating a DVD structure on an HDD"¡Xunder high-performance conditions, the software can combine the DVD image-building and burning process¡X"the limitation of the HDD seek time should be more sensitive. Hard drives are fast when used only in write or read mode; switching from read to write on different HDD locations can slow down the overall processing."

REAL HELP

One of the ironies of "real-time" performance working its way into the marketing hype surrounding so many competing software tools¡Xparticularly given that the tools' creators cite hardware factors as real-time rendering's real enablers¡Xis that nearly all of them claim that their real time is faster than the others'. Marketing-message ironies aside, what can an ingenious, enterprising software developer actually do to speed the encoding/rendering process?

"We improve the performance of our codecs by carefully optimizing them on the target processor architecture," MedioStream's Desproges says. "The easiest way to speed things up is to lower the quality. For example, the lower search range of motion estimation would require less processing power, but will result in lower image quality. Ultimately, other encoding techniques based on pattern recognition could be used (and still compatible with MPEG-2). Variable bit-rate is helpful as well."

According to Dazzle's Williams, accelerating the transcoding process via software "mostly involves finding more efficient block-searching algorithms and taking advantage of multimedia hardware capabilities (e.g., SSE) as much as possible to offload the CPU. Thus, the performance of an encoder can be very different on different processors of the same basic speed depending on whether that processor has been optimized for the multimedia extensions available."

Pinnacle's Lane says the first tack to take in enhancing rendering speed via software is "taking advantage of processor shortcuts, such as Pentium 4 optimization, which can shorten the encode time by 25%." Other strategies he suggests include "fast hierarchical motion estimation, more effi- cient algorithms for bit allocation, motion compensation, and block coding."

Sonic's approach, according to Mark Ely: "Speed the decoding and encoding through code optimizations. We have done a lot of this, and continue to do it to improve performance¡Xbut not at the expense of video quality. We will always err on the side of slower rendering if the end-product looks better."

TALE OF THE TAPE (TO DISC)

So, whose "real time" proved fastest in testing? In some instances, it was pretty close; in others, not close at all. One thing I learned about these products in the testing process is that there are five ways entry-level DVD authoring software can be optimized to get quick-and-dirty capture-and-burn done fast:

1. By integrating the three processes from the user's perspective, so that all parameter-setting and other user input happens before the capture begins and the next thing the user does is label the completed, ejected disc

2. By otherwise limiting the paths the user has to follow in the interface (not having preset defaults, requiring clicking to other windows to check parameters, requiring minimal or more-than-minimal menu creation, etc.)

3. By choosing lower-quality video encoding as defaults

4. Speeding the encoding process in "behind-the-scenes" ways as described earlier

5. Recording "on-the-fly" (while building the DVD image) on faster systems

As Dazzle's Marc Williams pointed out to me, there are too many variables here¡Xbit-rate, time, quality¡Xto enable a purely scientific test. His suggestion: "Test the 'default' encoding behavior of the applications as shipped and compare them for time/quality."

In part, that's what I ended up doing. One key factor made this approach both more effective and more satisfying: as far as I could tell, in all the products I tested, the preset default bit-rate was highest-quality 8Mbps, which made for more realistic comparisons and better results.

I also did some basic usability testing. Four of the tested products feature a capture-and-burn mode, though they all name it differently: PowerProducer calls it Quick Burn; DVD Complete calls it QuickDVD; MovieFactory calls it Direct to Disc; WinDVD calls it Direct Recording; MyDVD 4 calls it Direct-to-DVD. Only two of the tools tested, in my experience, effectively streamline the process from a usability standpoint¡X MyDVD and neoDVD¡Xin that they get all the user interaction out of the way before the capture starts, and then the software takes it from there.

Some of the tools tested also require some user input into menus, chapter creation, etc., along the way. In testing, I kept that at a minimum, though it was interesting to see how with minimal requirements fulfilled, the resulting discs matched up¡Xe.g., how between two discs created in roughly the same time, at equal encoding bit-rates, and with (apparently) equal video quality results, one might have a nice, cleanly organized menu with three or four chapters and the other no navigation whatsoever. For fairness' sake, we'll call that the subjective part of the test.

All the tools were tested with the same 10-minute DV clip, moderately motion-intensive with roughly a dozen scene changes. The clip was captured from a Sony DCR-TRV18 miniDV Handycam, at full DV quality and 720x480 resolution, NTSC, 29 frames per second. All tests were done on our in-house testbed PC, a 2.4gHz Sony VAIO with a 133mHz system bus, and 512MB RAM, using its factory-installed FireWire card and a factory-installed A04 DVD-R/RW drive with 2X TDK DVD-R media. Video was captured to the D: partition of the system's on-board 133mHz 120GB hard drive. Results were analyzed on a standard 4:3 TV using a Pioneer DVD player and from the A04 and a 16X LiteON DVD-ROM on Sony's SDM-X72 TFT-LCD flat-panel display, an XGA model driven via DVI by a Matrox Parhelia AGP card.

This is the story.

SONIC MYDVD 4

MyDVD 4 scored high marks for usability by consolidating capture-to-burn in a single uninterrupted process. What's more, it joined neoDVD (see later) in performing seven minutes faster than any other tool tested in accomplishing an identical feat.

The opening MyDVD 4 screen gives you three options: Edit Disc, Edit On Disc, or Direct to Disc. After you select the Direct to Disc wizard, the first option MyDVD presents is select menu style. Here you can choose "No menus¡Xjust play movie" or go with the default "Allegro" screen. Since it was the same number of clicks, I chose Allegro. On the next screen, you select record length (10 minutes). Next comes Device Settings Options: I went with the default, Best (8Mbps); alternatives were Better (6Mbps) and Good (4Mbps). Here I also took the single-click options "Disable preview option for better quality" and "Chapter point creation by time interval" (default: 3 minutes¡Xyou can also add chapters manually by hitting the space bar during capture, but that didn't seem in the spirit of the project). MyDVD took it from there.

MyDVD 4 delivered the all-around best direct-recording performance of all the tools tested. From the time I opened the program to the moment the completed disc popped out, MyDVD 4 ran a 19:06, which placed it second to neoDVD in overall time elapsed. In my book, this qualifies as "real-time" DVD creation, breaking the 20-minute barrier for capture + burn of 10 minutes of video.

What's more, MyDVD 4 overcame the 20-second differential between its running time and neoDVD's in the effectiveness of the disc produced. While video quality difference was indistinguishable, the MyDVD disc featured a clear, readable menu with four chapter points set automatically by time interval throughout the 10-minute span of the disc. [For more on MyDVD 4, see review, November 2002, pp. 46-49¡XEd.]

MEDIOSTREAM NEODVD

In terms of sheer speed and streamlining of the process, neoDVD took the prize. Keep in mind that all these tools do much more than direct recording, and MedioStream also strikes me as the overall best tool out there for the range of tasks performed by entry-level tools, even though I give Sonic the direct recording nod (in spite of neoDVD's slightly faster speed). As usable as any tool out there, thanks in large part to its sparse but effective neoTasks taskbar, neoDVD is slightly counter-intuitive when it comes to direct recording, in that it all happens under the "Create" taskbar, rather than the "Capture" taskbar, but once you get over that hump, it's remarkably smooth sailing.

Click Create in neoTasks (which appears when you open the software), select the wizard icon, and you'll bring up a screen that prompts you to make three choices, two of which have preset defaults: Select Disc Format (default: DVD), Video Quality (default: Best 8Mbps), and Viewable in. I went with the two defaults and NTSC (over PAL). Then you identify the video source (default: FireWire DV camera), set capture duration (10 minutes), choose menu style (quick drag and click), choose thumbnail frame (quick drag and click). Then it asks you where to write (default was my correctly ID'd A04 DVD-RW drive). Then neoDVD does the rest, all of which (including choosing all those settings) took 18:46. Except for Sonic, no tool even came close.

Video quality was great, as expected (again, Best was a wisely chosen default). As mentioned before, the Sonic disc edged neoDVD's because it let me make a basic menu without prolonging the settings part of the process (getting into that stuff with neoDVD would have taken too long). But menus are hardly a high priority in this type of application, and neoDVD was the clear winner in the speed race.

INTERVIDEO WINDVD CREATOR PLUS

InterVideo WinDVD Creator Plus is a nice all-around consumer DVD authoring tool and a capable direct-to-disc competitor. [See January Tools of the Trade, pp. 60-61¡XEd.] You can get down to business quite quickly in Creator Plus, but your input will be required in the later stages of the process more than in neoDVD and MyDVD. In the attractive opening window you select Capture, and after the software IDs the capture device, click record, and click stop when you're done. Then click Make Movie, and you'll bring up the Burning Wizard.

Here, as in neoDVD and MyDVD, the defaults gave me what I was looking for, which kept things moving (and kept the comparison consistent): for Format, DVD, 1 hour, which meant Best Quality and 8Mbps MPEG-2 encoding.

Creator completed the capture, DVD image creation, and burning process (plus minimal parameter-approving and navigation time) in a respectable 28:33, which is nearly 10 minutes slower than neoDVD and MyDVD, but certainly a manageable amount of time for getting highest-quality video captured and recorded to DVD-Video, all things considered.

SCM DAZZLE DVD COMPLETE

SCM Microsystems' Dazzle DVD Complete gets off to a quick start for direct-to-disc DVD creation and keeps things moving along from there. The opening screen is a "Select DVD Project File Wizard" that gives you five choices (which strongly hint at SCM's ambitions for the tool): QuickDVD, HomeDVD, BusinessDVD, HollywoodDVD, and Blank Project.

I selected QuickDVD and moved onto a new screen where I clicked "+" to "add movies," selected "From Video Capture" to identify the source, and clicked the red record button to begin capturing. After capturing for the proscribed ten minutes, I advanced to a screen which gave me scene-trimming options (usual scrollbar stuff), and clicked OK to move on to rendering.

From here on in, DVD Complete does a great job of providing stats on its activities: 6:07 for rendering (again, default bit-rate setting: 8Mbps), 61:1 rendering factor (interesting); then it built the "GVP" project file from which it would compile the DVD. Once the GVP is built (clock still running), it's a single click back to the QuickDVD Wizard, and another click to Burn DVD. After 2:47 of DVD image-building, burning begins. Producing excellent results once again (great-looking video, no menus or chapters), DVD Complete got the job done in exactly 26 minutes¡Xsix minutes off the 20-minute "real-time" standard, but again, not a lot of time to ask to crank out 10 minutes of playable DVD-Video.

THE OTHER ONES

Only four of the tools tested achieved anything close to real time. I defined "close" as at 30 minutes, which essentially means 2X real time for a 10-minute captured video file (10 minutes capturing time, and 20 minutes to transcode/render, build a disc image, and burn). MedioStream neoDVD and Sonic MyDVD 4 left the pack eating their dust, finishing under 20 minutes, while WinDVD Creator Plus and Dazzle DVD Complete came in safely under the 30-minute cutoff.

The others performed a good bit slower¡Xall over 40 minutes¡Xon multiple attempts with a system cleanly restored to pristine pre-test conditions. All those tools have much to recommend them as entry-level DVD authoring tools (see, for example, the Pinnacle Expression review in January Tools of the Trade), and nice, clean interfaces for direct recording, but I can't recommend them as real-time or near-real-time direct-to-DVD utilities.

Stephen F. Nathans (stephen.nathans@infotoday.com) is editor of EMedia.

Comments? Email us at letters@onlineinc.com.