Applications /

A card catalog for your multimedia

Repurpose your video assets to save money or make new revenue.

By Christine Perey
Network World, 09/10/01

As IP-based video and voice moves beyond executive speeches (also known as "glamour flicks") into training and other applications, corporations will begin accumulating streaming media content in various formats and places.

Unfortunately, most companies lack the technologies to turn their video archives into assets. Even with a rudimentary archiving system, the majority of a company's archived videos aren't systematically named, nor are contents easily determined by looking at a thumbnail. Some are segmented and synchronized with PowerPoint presentations. Only the most tenacious individual could recycle and repurpose them.

But products exist to solve this problem, so we set out to examine the options for a midsize corporation that wants to archive its digital video with the goal of reusing segments for new information campaigns, such as product launches. We tested Convera's Screening Room, eMotion's MediaPartner, MediaSite's Publisher Suite, and Pictron's Video Gateway, Audio Gateway and Media Gateway Search Server. We learned that there are more products available than hosted applications. Additionally, despite a depth in functionality, the products were easy to use - once they were customized for a particular corporate workflow.

Why IT should care
Considering multisite collaboration
The services alternative
How we did it
NetResults

We decided that Convera's Screening Room was the best solution for a company with divisions in multiple locations, where video was prevalent and the ability to search it would likely increase revenue. Screening Room is a comprehensive, well-integrated and scalable product with an open architecture that promises to continue expanding through Convera's internal development efforts and partnerships with third parties.

RFP responses

As part of our tests, we solicited responses to a request for proposal (RFP) to assess what vendors offered and the costs associated with their product or service (see How we did it). Companies responded as if in a competitive bid situation.

Pictron did not prepare a proposal, saying its package does not require professional services for installation and management. Its cost ($20,000 for its Video Gateway, Audio Gateway and Media Gateway Search Server applications) follows a standard software licensing model. If one were to buy four similar Video Gateway and Audio Gateway systems (one for each product division in the RFP we issued) and use a single Media Gateway Search Server, the cost would rise to $50,000.

EMotion's MediaPartner 4.0.2 provides a seamless environment for capturing, searching and collaborating with all popular media types. The company's proposal for our fictitious company's digital media ingestion and management project included word processing documents, Adobe Acrobat files, video and still images. While the proposal included integration with a third party's video capture and indexing product (MediaSite's Publisher Suite, one of the products we were already looking at), it differed substantially from those offered by MediaSite and Convera. MediaPartner's core technology focuses on virtual collaboration, annotation (on and around the video frames), and video creation-through-approval workflow notification, as opposed to the repurposing of video assets. MediaPartner also did not include its license cost in its RFP. It would be at least $15,000 for the capture component of MediaSite Publisher, in addition to the $100,000 price for MediaPartner.

The two most similar packages we tested, MediaSite's Publisher Suite and Convera's Screening Room, performed numerous steps, including media capture, indexing, searching and export to a Web server. The cost to purchase either product for an "unlimited number" of users was well over $100,000 (not including the cost of hardware for processing, storage, networking, digitizing cards and the license of an enterprise-grade Open Database Connectivity database). MediaSite's proposal for four Publisher Suite stations and one networked Video Application server, Application Developer's Kit, training and one-year maintenance agreement would cost approximately $160,000. Convera's equivalent system, with four complete Capture and Edit/Browse stations and one Video Asset Server, would cost more than $250,000. A scaled-down implementation of Screening Room with a single (centralized) media capture and editing client would cost $120,000 less and still permit full asset searching and viewing by users anywhere in the company's enterprise network via a Web browser.

To have these investments amortized as quickly as possible, the product should change the way at least some people in your company do business. Risks associated with workflow modification or process re-engineering are hard to measure, and the potential return on investment is often less than predicted. However, recycling and repurposing existing video, while making it a more integral part of the communications process, should cost less than the cost of recreating the content from scratch. Additionally, it will have a more lasting positive effect on remote audiences than text and graphics alone.

Enterprise-worthy?

The first characteristic we assessed in our tests was how enterprise-worthy the products were. We were looking for the following:

1. An ingestion system, consisting of metadata capture and video capture.

2. A sufficiently robust relational database system with a vendor-provided application that includes keyword searches, time-based searches, visual searches or searches on other metadata (such as annotations) from anywhere on a corporate network.

3. A clip viewing/publishing environment and a Web server (for end users who only have Web browsers).

4. A streaming media server (if using a format that isn't integrated with the Web server), and in some cases, a high-capacity, high-accessibility storage system.

These features were present in MediaSite's and Convera's products. Convera's implementation appeared more seamless, although analogous required parts and additional "modules" were available in the MediaSite setup we evaluated. EMotion's product, at least the version we saw, did not permit us to conduct all the tests we set out to perform (such as automatic scene change detection and speech-to-text transcription during content capture), because there were version incompatibilities between the acquisition system (MediaSite Publisher Indexer) and eMotion's MediaPartner 4.0.2. This said, MediaPartner's tools are highly appropriate for collaborative processes such as video review for post production, annotation and obtaining client or management approval on finished content (see "Multisite collaboration," page 76).

The configuration of Pictron's suite we tested offered the first and third components listed above. However, we think Pictron's suite is best for a small company or a workgroup. The suite ships with Microsoft Access for its database, and we feel that a midsize or large corporation would quickly tax Access' capabilities. Pictron's architecture can accommodate more robust databases, such as Oracle8i or Microsoft's SQL, but we weren't shown that implementation. Even with Access, Pictron's Video Gateway succeeded in completing our ingestion and searching tests, offering a compelling price/performance point for companies that don't anticipate a high degree of enterprisewide workflow integration.

Automation and accuracy

Our next objective was to see how well the automated video indexing systems worked without any special configuration, design or speaker training. Because enterprise media doesn't have closed captioning as is often found in news media, we suspected results would vary with the source. We chose to use compressed digital files (.avi) as input because we didn't want to have analog-to-digital boards in each workstation. Vendors warned us that most companies they work with don't begin with digital archives (so we were deviating from a common scenario), and that the audio would have less-than-optimal clarity and fidelity because of compression and sampling of the digital files. We also realized the technology we were using was optimized to automatically transcribe broadcast news content.

Convera, MediaSite and Pictron all license IBM's ViaVoice for their speech-to-text transcription engines. EMotion's package can be built on any of the video capture tools that in turn have the option of licensing IBM's technology. The text generated by the engine was unusable on all three content sources we used. One source, a 4.5-minute marketing video, had a background music track that interfered with the systems' understanding of the narrator's speech. A professionally produced, quarterly corporate earnings presentation without music also produced aberrant results.

We were informed by Convera's team that if a transcript was made of a speech, the words would be synchronized automatically with the video in software. Additionally, Convera said its customers could have content transcribed (at a rate of approximately $300 per hour of digital video) by its partner, WordWave. Then, using an import utility from Convera, you could synchronize the file back with the video, producing a cost-effective and accurate alternative to the automated or internal transcription options. Convera had this done on some of our test content prior to coming to the test facility and demonstrated the results of partnering with WordWave.

The automated, video scene change detection features were comparable in the three capture systems tested, as was the ability to have the software create thumbnails, also called key frames, when there were important video events (such as fades, transitions and lighting change) and/or at designated time intervals. However, we concluded that meaningful indexing of content and creation of short, logical segments from thumbnails for searching remains manually intensive.

In our tests, all three indexing engines performed video and data analyses in real time on the server configurations we were provided (in other words, we did not detect any difference in capture speeds and indexing efficiency). We suspect that as the number of users increases and the server is asked to complete more tasks simultaneously, the time to capture and index a file will increase.

Ease of use

We found the graphical user interface in Pictron's 3.0 release very attractive and the most "Web-like" in design, although some icons did not follow standard Windows conventions. For example, an "x" normally stands for remove or "unavailable," but in Pictron's clip assembly interface it stood for "OK." The sparse, text-search screen layout leaves a lot of room for customization by an administrator, a good characteristic in our estimation, as long as you have the skills in- house to take advantage of it. The search results from the Pictron Media Gateway Search Server were easy to read. All vendors' search results pages also used an embedded player window for returning and reviewing the video results.

Since the other packages offer a high degree of flexibility (where many options are offered at the same time), we found that some screens appeared crowded (available options were competing for screen real estate). Fortunately, you can customize most of the layout in the search and viewing mode using a professional services group, a reseller, or in-house engineering with the vendors' respective application developers kits. MediaSite and eMotion also address this effectively by dividing up the features and workflow into component applications.

Our favorite user-friendly tool was Convera's Screening Room Capture system configuration wizard. The wizard guides users through potentially complicated decisions about media capture; file management and posting to content delivery network; display; logging; transfer of metadata to database; searching; and publication. The tight integration of the different parts of the application will likely appeal to users who are responsible for a complete workflow.

In contrast, MediaSite's Publisher Suite has five functional subapplications (Indexer, Editor, Post, Database Creator and Database Tool box). While we were impressed with the applications' functionality and stability, we felt that this architecture could be confusing to occasional users because different steps in workflow were handled separately. We were surprised that MediaSite didn't provide a progress bar to tell us how far into the capture we were, and nothing in the interface indicated when the Indexer had finished. Convera's status window had a progress bar, yet the application also failed to report to us that it had successfully published when we posted to the Web server.

When in indexing and edit views, all systems had intuitive screen composition. Thumbnails created during indexing are organized in rows in the upper right- or left corner; text in lower right- or left corner; and the original media viewing screen in a separate window. From an asset browsing and organization point of view, the default is to use Windows directory conventions. Pictron's layout looks a little more like a Web site navigation bar, reflecting its recent development.

We liked that Convera and MediaSite offered wizards to schedule automatic captures from user-designated sources. As in other Convera applications, Scheduler's dialog boxes and menus conform perfectly with Microsoft Windows GUI conventions, with tabs, pop-up menus and check boxes. Pictron's Media Gateway Director, scheduled for release later this year, will also permit scheduling of remote tasks in multiple servers. Although we didn't use them sufficiently to rate their accuracy and clarity, all four products come with online help and print manuals.

Posting, publishing and versioning

When you post or publish to a Web server, the results are visible to someone using a browser. In contrast, when a video indexing system posts, it is transferring information to the associated database in appropriate fields. Posting, editing metadata, publishing to a streaming server or Web server and facilitating access to media assets (allowing search on key words, for example) are steps that may or may not be performed by different people at different points in time. An administrator can also automate some of these steps.

Convera's SR Capture initiates all post-capture processing as defined during system setup. In MediaSite we found the manual process of posting slightly less intuitive. It appeared to have extraneous steps that may be valuable for a configuration we weren't seeing. We also understand that if the workflow were segmented and assigned to different people (with separate administrative and logical activities), it might be advantageous to have an architecture such as what's found in MediaSite's Publisher Suite and Video Application Server.

Once metadata and video assets are posted, employees (corporate communications or product marketing managers, for example) should be able to access all the databases, preview the content, select certain assets and push out any number of files to an appropriate page based on the target audience. Although not an inherent part of the edit or browsing views, Convera's Video Asset Server and MediaSite's Video Application Server have administrative tools and APIs that can be customized by Web programmers to generate templates into which the thumbnails are published.

Conclusion

Overall, we were very impressed with the degree of maturity and stability in products we tested from Convera, MediaSite and eMotion. This said, we want to reiterate that the science of managing video as a corporate asset is young and very rapidly evolving. We expect new features designed for corporate content will be available as modules or "objects" within the present architectures. We also expect that some of the services, such as the one hosted by Pictron, will be popular among customers that want to test the waters but not risk drowning in complexity.

Considering the sizable investment necessary to bring these systems in-house, we recommend that anyone planning to make video content searchable identify multiple target applications and not be afraid to change internal workflows to more fully leverage existing video assets.

Editor's note: SonicFoundry announced its intention to purchase MediaSite on Sept. 7, 2001. Read the press release.