A Buyer's Guide to Workflow Systems
Online video workflows are complex and tricky to pin down, but focusing on a few key areas can make the entire system much more manageable.
So you've put together a script, shot and edited your video, and received approval from the client, and now you're ready to send it out into the wilds of the internet. There's only one problem: What format (or formats) do you recommend to the client?
The client, it turns out, wants you to try many different formats and wants someone to check the quality on each one, to determine which might be best for each of the client's intended audiences. But the client's not particularly keen on paying for someone to babysit the transcoding system for all those hours.
This particular Buyer's Guide won't give insight into the best format, codec, or even the best media server-all of those topics are discussed elsewhere. Instead, we'll focus on the questions around systems that can automate parts of the overall workflow.
When I started a side project, the Workflowed blog, back in 2009, little attention was being paid to the distribution side of the workflow equation. After all, we were just beginning to explore delivery solutions beyond the DVD, and we were experimenting with early device-based delivery with offerings such as Netflix and Apple TV. Several companies were beginning to tackle the bigger picture of automating both transcoding and quality control, albeit in customized enterprise solutions, but the "typical" workflow wasn't close to being defined.
Even today, workflows are tricky things to pin down: There are often many moving parts and sub-components to an overall media workflow-such as the interaction of transcoding and the editing process mentioned previously-but there are a few key areas to consider that can make this scenario a bit more manageable.
This question is key when weighing the decision to move beyond a simple transcoding system to a more robust workflow automation solution. For many agencies, and a few key broadcasters and enterprises, there is a consistency to the workflow that lends itself to automation solutions. For the majority of scenarios, however, the thinking around workflows is somewhat esoteric and not fully formed.
One of the first steps is to determine whether the workflow will involve either an on-demand or live component, or a combination of the two. In this Buyer's Guide, we're focused mainly on file-based workflows for on-demand content delivery, but there are several broadcast-centric solutions for live workflows that involve the type of real-time transcoders and format convertors mentioned elsewhere in this year's Sourcebook.
The short answer is "no," but the reality is that many workflows share commonality. The best process is a systematic "trial and error" approach, based on a foundation of best practices available from particular workflow automation product manufacturers or from professional services and consulting organizations.
One step in this process is to determine whether profiles will be based on particular client needs, or whether it is possible to reuse profiles across clients. As such, until a company defines the primary steps in their workflow, it's best to stick with transcoding solutions that are flexible for use in a variety of workflows.
Many of these transcoding solutions will offer the ability to define profiles that can then be replicated and slightly changed to accommodate a wide range of end-user devices. A few transcoding systems can even be upgraded to workflow automation systems or can act as a plug-in to a larger workflow automation solution. For both of these approaches, the experimentation allows you to use the basic system to define segments of the workflow that can then be added into the larger workflow.
When I moved from film production to video production, I learned about a vexing problem called tape dropouts. It didn't seem to be that big of a deal when it came to shooting on hideously expensive Beta SP tapes, but the reality of tape dropouts struck home one day when a key client brought in a VHS copy of a finished videotape-one of 60 we had duplicated for the client-that was clearly showing signs of tape dropout.
The only sure-fire quality control solution back then was to have someone view each tape. Alternatively, we could also run each duplicate through an expensive machine that checked for a "tolerable" level of tape dropouts, or we could have someone rotate through viewing of the output of several duplicating decks. None of these were good solutions, as the expense outweighed the benefit- unless it meant losing a key client, which would only become obvious in hindsight.
Fortunately, today, there are a few more ways that quality control can be managed within lower opportunity cost parameters.
One approach involves the use of a predefined quality control test bank-often based around peak signal-to-noise ratio (PSNR)-that "views" the finished transcodes for significant variations in quality from the original. PSNR has an Achilles' heel, though, in the fact that its mathematical equations don't mimic the human eye's perception of quality.
We've discussed this in articles on Streaming Media.com and in last year's Sourcebook, and the industry still needs to address the shortcomings of PSNR, perhaps by using a more robust quality algorithm or by tying the PSNR step in to a larger quality-control workflow.
Another approach to quality control harks back to the "spot check" approach, with a twist: Some workflow automation solutions allow retranscoding of particular sections of a file, without requiring the entire transcode session to be rerun. This works best for full-length transcodes that are already sectioned up-think VOB files and MPEG-2 Transport Streams-but can also work for today's adaptive bitrate files that are also being packaged and segmented as part of the transcode workflow.
A variation on this approach is a profile-based quality control, where the system notes which profiles fall outside of a particular set of parameters-bandwidth spikes, audio pegging, etc.-and then "fails" those particular profiles. A robust workflow automation solution will allow just those profiles to be retranscoded, after tweaks are made to address the out-of-bounds parameters, without requiring all profiles in a workflow session to be transcoded.
It's not enough just to create a workflow solution and then walk away from it. Like any good engineering solution, workflow automation is designed to free the operator from menial tasks so that they can move toward value-added tasks.
One of those tasks is assessing whether the workflow solution needs to expand, to encompass more of production or post-production processes. Another task is to consider scaling the solution beyond an enterprise workgroup or business unit or even beyond the physical building or campus.
As such, keep in mind the power of hybrid workflow automation solutions, which take equal advantage of the local workstation, onsite server clusters, and cloud-based workflow components. Several workflow automation systems have load-balancing capabilities to spread the work across multiple machines and a few are beginning to introduce auto-sharing databases that provide the advantage of seamless scalability.
A Worthwhile Investment
We've just scratched the surface of today's workflow automation options. Other areas to consider are the reuse of decoded content across multiple servers or workstations, as well as review and approval processes. In addition, automating delivery of approved content to content delivery networks (CDNs) is a key feature in both on-premise and cloud-based workflow solutions.
With any robust automation solution, the amount of human interaction with the system will be front-loaded toward tweaking profiles and quality control parameters. This workflow optimization can take on the form of programming software, or it can be geared toward changing human behavior to better suit the overall creation and delivery of key content in a more timely and consistent manner.
Yet the time spent thinking through all the steps in a workflow should yield significant time savings and lower overall production and distribution costs. If automation can free up resources to enhance the creation of content, it will be time and money well spent.
4 Questions to Ask:
- Is my workflow defined?
- Is there a best process?
- Is quality control important?
- Can my workflow expand or scale?
This article appears in the forthcoming 2013 Streaming Media Industry Sourcebook.