2010 Best Workflows

A bit longer in the making than anyone would like, but full of good information to set a baseline for live encoding and file-based transcoding workflows. We're happy to present the 2010 Best Workflows report, another in the Transitions in Technology white paper series.

Within the next few days, we also hope to be able to reveal details from a few participants. We'll keep you posted.

[Update: Elemental Technologies chose to publish their findings, which can be found at this link as the 2011 Q1 Best Workflows report, containing the same information as the 2010 Best Workflows plus additional information specific to the Elemental outcomes. Enjoy!]

Has Fujitsu Abandoned the Macintosh Platform? ScanSnap Snow Leopard Issues

Fujitsu had a good thing going, with the ScanSnap for Mac offering one of the best scan-to-PDF options.

But like all good things, this one looks like it's coming to an end, as ScanSnap on the Snow Leopard platform is walking around gingerly on only two paws.

Consider this:

1. The S1500M is the only current Mac-based ScanSnap scanner; all the others (S300M, S500M, S510M) are on Fujitsu's discontinued list.

2. CardIris 3.6, which ships with the S1500M, is not compatible with Snow Leopard (OS X 10.6). CardIris 4.0 is compatible, but you don't receive that version when buying an S1500M

3. Adobe Acrobat is now at version 10 with Acrobat X. Which version ships with the S1500M? Acrobat 8. Yes, that's right, the only Fujitsu Mac ScanSnap has a version of Acrobat Pro that's TWO versions old.

When I asked Fujitsu about upgrading to Acrobat 9, six months after it launched, I was told that what you buy is what you get. Meaning that anyone buying an S1500M today will get Acrobat 8.

Fujistu is clearly choosing to abandon the Mac platform, as they've not made a ScanSnap unit that's Mac compatible since 2008.

How's that for progress?

PDF viewing in Apple's Preview looking fuzzy? Welcome to Acrobat X on Snow Leopard

I'm working through a few workflow scenarios with Acrobat X's scanning and optical character recognition (OCR) for an upcoming comparison study. Along the way, though, I have two quick insights in to Acrobat X.

First, it does a great job of shrinking the file size automatically after the text recognition is run. In many cases, I don't even have to run the "reduce file size" feature to bring multi-page documents to a manageable file size.

Second, the issue that cropped up under Apple's Preview in Acrobat 9, after I'd applied the "reduced file size" to shrink the file size, remains. The image will look blocky and jagged, as if a great bit of information has been abandoned.

I initially chalked the issue up to an error in the file reduction algorithms, but the advent of Acrobat X—with its better-than-average standard compression—brings the issue back to the forefront. Even without running the "reduce file size" feature in Acrobat X, any file that's been OCRed will appear jagged and blocky within Apple's Preview application.

Researching the topic online didn't turn up any clues, probably because Acrobat X is so new; since I couldn't ignore this issue, as it affects all the scans I was applying OCR to, I turned to my contact at Adobe's PR agency to get insight in to the issue.

It turns out that Adobe X's new rendering engine may steal a few pages from Acrobat 9's file size reduction score. In doing so, this presents a potential rendering issue in Apple's Preview.  According to my contact:

Acrobat X's new scan compression technology divides the image into 3 layers - Background (BG), Foreground (FG) & Mask. BG, FG images are highly down sampled while mask is kept at a higher resolution (to maintain text readability). 


The layering makes sense, as Adobe has always had the ability to choose Image-Text (where the image overlays the underlying OCR text) or Text-Image (where the text attempts to lay out in a patter closely resembling the image, but the text is the top layer).

I've always opted for Image-Text, as it allows the human looking at the document to read what's actually on the page, should he or she find the OCR text they copied from the PDF a bit, well, lacking.

My contact went on to provide some reasoning behind the miss-match of Preview and Acrobat Reader X, the latter of which seems to display the images in a much higher quality output:

Here is our hypothesis on the reason of low quality rendering by Preview. In order to render a page, it down samples mask to the resolution of FG (or maybe BG), so it loses on text crispness or quality that a high resolution mask provides. Adobe Reader/Acrobat, on the other hand, up samples images to the highest of the resolution of BG, FG and mask to get the rendered bitmap.


Is this the reality? I guess we'll have to see whether Apple issues an update to Preview in the near term. If not, I'm stuck either suggesting that every client upgrade to Acrobat Reader X, or choosing to completely forego the workflow of using QuickLook to view text-heavy OCRed documents.

Transitions In Technology Series: Juniper Media Flow

Over the past few months, we've been hard at work on a workflow article on Juniper's Media Flow technology, which the company has been integrating since the acquisition of Ankeena.

This paper is called "Refining Media Delivery: Intelligent Caching Comes of Age" and covers a variety of aspects around intelligent media caching. This is the first paper in the Transitions in Technology white paper series, a series of commissioned documents that leverage the Transitions' team's experience in the marketplace to tell a compelling story. 

In this case, it's a story that we've followed for several years, as Ankeena announced Media Flow Controller. We think the acquisition bodes well for Juniper's move in to the scaleable media marketplace. 

Here's the link to the white paper. Enjoy!

Does Adobe Elements 9 Plays Well With All?

Adobe today announced its Elements 9 consumer image- and video-manipulation products (editing sounds so 1990s) in the form of Photoshop Elements 9 and Premiere Elements 9.

Over the past few releases, Elements has begun to move toward parity on both the Macintosh and Windows platform, but the Mac version had always lagged behind.

It appears the days of second-class citizenry for Mac users may be over, though, as Adobe lists an equal feature set between the Mac and Windows versions of both Photoshop Elements and Premiere Elements.

Premiere Elements, in particular, now has the sharing and access to the Plus service, which gives up to 20 GB of storage for approximately $50.00 per year.

In addition, Premiere Elements 9 addresses the ability to use tapeless workflows—from Flip cameras to D-SLR consumer and pro cameras—a feature that's only recently been added to Adobe's flagship editing tools, Adobe Premiere Pro, as part of the much more expensive Creative Suite 5 software bundle.

Speaking of pricing, that has also dropped for Elements 9 bundles with Amazon reporting the combo pack—Photoshop Elements and Premiere Elements—that's pictured below clocking in at $149.00 but with a rebate that drops the price to $119.00 (after waiting for several weeks to receive the mail-in rebate check, of course).

We look forward to testing the two products, on both platforms, to compare the feature sets, before the product release on November 1, 2010.  [Update: Adobe's PR reps say that the product is available now, even though Amazon still lists the availability at November 1]

SNAFU - Toast 10 Titanium Pro's Addition Problem

For Macintosh-based content creators, there aren't many options for Blu-Ray disc creation: Toast 10 Titanium Pro is the least expensive alternative that has an option for Blu-Ray conversion and burning. It clocks in around $110 on Amazon compared to several hundred dollars for Adobe Encore (and a downloadable version of Toast 10 Titanium Pro offered by Amazon is even less expensive).

Yet, for all the Toast goodness we've seen from Roxio over the years, this Blu-Ray version of Toast 10 Titanium is a dud: it can't calculate storage space in any meaningful way.

I found this out the hard way, having chosen the "Blu-Ray Video" setting and then spending 8+ hours digitizing standard-definition (SD) miniDV content into Toast 10 Titanium Pro.

Once my content was all digitized, I let the system choose its own automatic encoding, which yielded a pleasantly spacious additional 777.1 MB left on the theoretical Blu-Ray disc:

Except that's Toast Math, which means it bears no semblance to reality. While the Toast calculation tells the user that only 22.55 GB will be placed on the disc, the reality is that this number already exceeds the amount of space on a single-layer Blu-Ray disc!

How's that? Don't Blu-Ray discs have 25 GB of unformatted space (or 23.31 GB of formatted space)? Yes, gentle reader, you are correct. But in the world of Toast Math, 22.55 GB is greater than 23.31 GB.

To fully understand the equation, you must allow your computer to crunch data—lots of data—for more than 36 hours. At the end of this time, your computer will then generate the following error:

Cross-check the Toast Math: Toast calculates 22.55 GB out to be equal to 23.76 GB. The latter is the amount that Toast says it needs to burn the same disc that started with the original amount of 22.55 GB on disc (and 777.1 MB of free space!)

I thought this was an anomaly, so silly me, I re-ran the same scenario for another 36 hours, and used a different Blu-Ray (BD-RE) writer.

Guess what? Same exact issue.

Then,  foolishly, I tried making a stand-alone Toast image (ending in .toast) and burning it via the Copy "Image File" option in Toast 10 Titanium Pro.  The .toast file ended up at almost 25 GB, almost 2.75 GB higher than the original Toast calculation of 22.55 GB).

Yet, when I chose "Image File" under Copy, guess what happened? You are right, gentle reader, I received an error telling me that 23.76 GB of disc space was required on the Blu-Ray disc, the exact same amount that Toast 10 Titanium Pro had calculated in error during its initial bad math day (we're now over 100 hours into testing Toast Math).

What about just removing one of the miniDV files from the Toast layout?  That seemed like a natural conclusion, but due to Toast's automatic encoding, it's not so simple: rather than just taking away the 3.1 GB +/- for each of the miniDV tapes, Toast 10 Titanium Pro does another round of Toast Math, and comes up with less free space on the Blu-Ray disc than before.

Perhaps I don't understand new math, or Toast Math, but apparently the re-calculation is based on Toast 10 Titanium Pro changing the compression rates to "fit" on to the Blu-Ray disc.

In theory this makes sense, since fewer minutes of video being compressed to the Blu-Ray disc could yield a higher-bitrate-per-minute scenario. But not in Toast Math.

In the world of Toast Math, 1+1 doesn't equal 2, 3 or any standard calculation.  It equals Toast Math.

Please, Roxio, teach your program to add before sending it out in to the world!

Systems Integration for the rest of us

Preparing to attend a show for systems integrators in Las Vegas this week, I spent part of today covering Steve Jobs' WWDC keynote address for StreamingMedia.com.

Read the details on iPhone 4 and my take on Apple's attempt to outgoogle Google with iAd and HTML5 lip service on the streaminedia.com site, but pause first to reflect on these excerpted comments that Jobs made. . . .

"Apple is not just a technology company," said Jobs. "It's the marriage of technology and the humanities that distinguishes Apple."

Echoing sentiment I've heard from Sling's CTO Bhupen Shah, who reminded me years ago that the ease-of-use for compelling products is all about the hardware and software working seamlessly together, Jobs drove the point home using the example of the new iPhone's second camera.

"On iPhone 4, it's not just a front facing camera: it's a front facing camera and 18 months' worth of work to come up with software that you'll never even notice when you want to place a video call," Jobs said.

"It's a complete solution so all of us don't have to be system integrators."

iPhoned

Why Partner? Edgeware and MediaMelon

As I've been working through partnerships between various professional transcoding and live encoding manufacturers, so that two companies can provide a joint testing solution for both areas of our comparative testing, I've been struck by the continued uptick in joint solutions.

One such joint solution, which I wrote about today for StreamingMedia.com is a combination of edge caching devices and deterministic middleware, with the two partners being Edgeware AB and MediaMelon, Inc., respectively.

You can read the technical descriptions of the joint solution at StreamingMedia.com, but here's a brief overview of the two companies:

Edgeware was formed about 4 years ago by networking and video professionals, who wanted to solve the challenge of addressing next-wave internet traffic. They felt much of the traffic going forward would be video, and they chose to design an appliance with caching ability. Since this first-generation 1U server used solid-state memory and had an FPGA for programming, IPTV was a natural target market. Edgeware's appliance had a Linux derivative with their own home-grown UDP stack and  file systems, and Nokia, Alcatel, Siemens all sell the appliance as solutions to large telecoms who want to offer walled-garden IPTV services. 

MediaMelon is backed by founder and chairman of Macrovision, with its main office in San Francisco and primary R&D in India. The company's deterministic algorithms, backed by significant viewer experience data, allows MediaMelon to offer its customers - many of which overlap the Edgeware customer base - a set of differentiated service including quality of service for UDP, TCP. MediaMelon, whether through its MediaMelonDirect, for media customers, or its MediaCloud SaaS / hosted solution for telecoms, ISPs and MSOs, leverages real-time viewing quality information and then serves up chunks of streams from a variety of locations across a CDN.

So when Edgeware wanted to move beyond walled-garden IPTV appliances and into the realm of TCP delivery of "web TV" on second-generation appliance, using RTSP and chunks of adaptive bitrate content, the two companies had complementary technologies. 

Yet, why did the Edgeware-MediaMelon joint solution make sense for customers and integrators, especially since both companies had a slightly different approach to addressing the joint solution? 

"From Edgeware's perspective," said Jon Haley, Edgeware's VP of  business development for WebTV and Over the top (OTT)," our VARs and system integrators are seeing huge demand for building out CDNs in local service provider markets and then adding in OTT services. The MediaMelon federated model really is unique, based on the quality metrics from a user's perspective (QoE), and matches nicely with our quality philosophy of pushing content further out into the network via our boxes."

In other words, for markets like Europe, which are highly fragmented due to language and/or incumbent state-run service providers, the joint selling of a federated model couldn't be accomplished just by Edgeware.

"It really is a natural partnership," said Kumar Subramanian, founder and CEO of Media Melon, "making it easier to jointly provide value to service providers who can quickly deploy a full solution of Edgeware boxes along with our SaaS hosted middleware. The value proposition of offering a large footprint due to the federated model is a significant solution to a significant problem."

Test Criteria - Professional Transcoding and Live Encoding comparison testing

As mentioned yesterday, two additional companies expressed interest in the comparative testing for professional transcoding and live encoding workflow systems: Sorenson and a new player in the space, Octasic (or 8 ASIC).

While neither of the companies has enough of a total solution to compete in this year's testing, both may have solutions in the near- to mid-term that would fill the gaps and allow them to compete during our second annual test, to be held in 2011. That still leaves 13 companies, including Ateme, Envivio, Harmonic, Inlet, Media Excel, Telestream, who are weighing their solutions against the test criteria.

What are the testing criteria?  Here's a generic overview of the way Transitions will test the various hardware-software solutions:

1. Transcoding: these tests come first, as a way to establish a baseline of quality, with speed secondary but almost equally important.

Here we'll look at things such as failover, including job completion and workflow steps; ability to transcode a single file to multiple adaptive bitrates.

On the speed side, we'll look at speed to transcode files from M2T (MPEG-2 TS), H.264 and another format back and forth to each of the other formats; whether the solution is capable of simultaneous transcodes; and we'll judge the speed of transcoding multiple files to multiple adaptive bitrates.

Resolutions to transcode to will be: 1080p, 1080i, 720p, several web and mobile rates.

Finally, we'll see the number of concurrent file-based transcodes that can occur, based on three workflows: Real-time, Faster Than Real Time, Real-Time xN

2. Live Encoding: after transcoding, we'll turn the testing criteria on its head and look at what each solution can do within a set period of time - specifically within the confines of live encoding testing.


Both tests will also assess performance, throughput and ease of use in setting up a variety of workflows, taking a holistic approach to pro transcoding and live encoding workflows. For instance, for clustering, we'll look at initial setup (3 total units, plus mgmnt node and 5 total units, plus mgmnt node), replacement of individual node(s) or the management node.

Part of the testing will take place in our lab environment, to allow for a common set of test criteria in a typical workflow. In another part of the testing, we'll jointly assess management, clustering and fail-over response, in your lab or quality control location – recommended, unless you want to ship a hefty rack of gear.

To keep the tests objective but still cover the testing time, a sponsorship model is being used. We've found with large groups of test candidates that a sponsorship with equal amounts paid by each sponsor - so that no one company dominates the report - works best.

That's the model we'll use for these tests, and more details on the approach, as well as examples of previous multi-party tests and test methodologies have been made available on request to all invited companies.

It continues

Two more companies have expressed interest in being included in the transcoding and live encoding comparison testing: Sorenson and a brand-new company in the space. More on that tomorrow.