By: Fereidoon Khosravi
We have achieved something that has been near and dear to my heart for a couple of years, as it relates to our Cloud-based QC capabilities. And that is how we extended the capability of our Quasar native-cloud QC service to the level that I don’t believe anyone else has actually reached! And we call it “Quasar Leap”.
When we started the development work on Quasar®, our native cloud QC service, the goal was clear. We didn’t want to just take our popular on-premise QC software, Pulsar™, run it on a VM and call it “cloud” QC. We made the deliberate decision that while we would use the same core QC capabilities of Pulsar, we would build Quasar architecture from grounds up, to be a ‘native’ cloud QC service. And we did accomplish that by being the first cloud-based QC to legitimately call ourselves ‘native’ cloud. And the phrase ‘native’ cloud meant capabilities like microservices architecture, dynamic scalability, regional content awareness, SaaS deployment, usage-based pricing, high grade content security, and of course redundancy.
But we wanted to go even further. And that was when the project we code named ‘Quasar Leap’ came about. To borrow and paraphrase from one of my all time favorite TV shows, Star Trek, we wanted to “take Quasar to where no cloud-QC had gone before”! (Those of you who are Star Trek fans know what I am talking about!).
Quasar was already able to process 100s of files at the time, but the goal of ‘Quasar Leap’ was to show that Quasar can process ONE THOUSAND files simultaneously! Of course, anyone can claim that their solution is robust, scalable, reliable, etc, but we set out to actually do it and then record it to prove that we did it!
This was not a marketing ploy, although to be honest, I knew there would be great appreciation and name recognition telling our customers and prospects that we can QC 1,000 files simultaneously. But there was a practical and quite useful benefit of doing so. After Quasar’s initial release, we found out when we started to push the boundaries of how many files Quasar could process simultaneously, there were some practical limitations to our architecture, even though we were already way ahead of our competition. And while we could easily process a few hundred files at the same time (more than any of our customers had needed), when we tried to push beyond that, the process could break down, and impact the reliability of the overall service.
So because of project ‘Quasar Leap’, our engineering team took a very close look at various components of our architecture. And while I am obviously not going to give away our secret sauce (!), suffice to say, they further enhanced and tweaked various aspects of our internal workflow to remove any bottlenecks and stress points, to make Quasar massively and dynamically scalable!
And then we decided instead of just ‘saying’ that we have the most scalable native-cloud QC solution, we would ‘actually do’ it and record it!! And I can now tell you confidently and based on the recorded video that we have actually done that! That, we actually submitted 1000 30-minute media files, watched our Quasar system dynamically spin up 1000 AWS virtual computing units (called EC2s in their terminology), and process (QC) those 1000 files simultaneously, and then spin down those EC2 instances once they were no longer needed.
So while we recorded the event, 30,000 minutes (500 hours!) of content was processed in less than 3 hours which even included the time of spinning up of the EC2 instances! To put things in perspective, 500 hours of content, equivalent of approximately 330 movies, or 10 seasons of 7 different popular TV Sitcoms, was processed just shy of 3 hours! To say it differently, with our massive simultaneous processing capability, approximately 160 hours of content can be processed in one hour!
If you say to yourself “that is great, but who has that much content that they need to process them that quickly”, I have an answer! Actually a three-point answer:
- You will be surprised how many media companies have PETA bytes (that is with a “P”!) of content sitting in cloud storage! They face the daunting task of managing a cloud-based workflow to monetize that archived content by restoring, ingesting, transcoding, and ultimately delivering that content to audiences. And one step that is naturally important in this workflow is ensuring that all of that content goes through content validation at various stages before delivery to the end user. And that is where this massive simultaneous QC processing ability of Quasar will be much needed to minimize delays in this effort.
- Some of our customers get content in bursts with strict delivery timelines. Ability to process that burst of content immediately offers significant business value in addition to workflow efficiency.
- And let’s not forget the main gain from the ‘Quasar Leap’ project, which was the behind the scene tweaking, in some cases revamping, and enhancing of our underlying architecture. And that has resulted in a solid platform, which will benefit ALL of our Quasar SaaS users, whether they have 100,000 files or 100 files, or even a few files! It ensures reliability, scalability and confidence in that they can rely on Quasar to meet their QC needs regardless of their normal volume or any sudden increases (bursts) in their content flow due to an unexpected event or last minute request.
All that effort for ‘Quasar Leap’ by our talented and dedicated development team, conducted during the challenging time of the pandemic, is finally complete! The new release of Quasar with all the architectural changes resulting from the ‘Quasar Leap’ project has rolled out.
According to Tony Huidor, SVP, Products & Technology at Cinedigm, a premier independent content distributor, a great customer of ours, and an early benefactor of ‘Quasar Leap’: “Given the rapidly growing volume of content that uses our cloud-based platform, we needed the ability to expand the number of files we need to process in a moment’s notice. Quasar’s massive concurrent QC processing capability gives us the scalability we required and effectively meets our needs.”
And now ‘Quasar Leap’, giving us the ability to massively scale up our simultaneous processing capability, is ‘live’, and “our Quasar native-cloud QC has gone where no other cloud-QC has gone before”!
Learn more about our Quasar capabilities here, or contact us for a demo and free trial!
And as Mr. Spock would say: “Live Long & Prosper”!
By: Fereidoon Khosravi
In my last blog, which coincided with the official launch of CapMate, our Caption and Subtitle Verification and Correction platform, I gave the background on how the concept of CapMate came about and, at a high level, what the capabilities are that it brings to the table. Here is the blog, in case you need a refresher! It is now time to dig in a little deeper into what CapMate can actually do and why we think it will add great value for any organization that has to deal with Closed Caption/Subtitle files.
We had many conversations with our customers and heard their concerns about the issues they ran across when processing or reviewing captions, and why the closed caption verification and correction is a slow and time-consuming process. Based on those feedbacks, we derived a list of key functionalities which would allow them to reduce the amount of time and effort they regularly had to spend in verifying and fixing the caption/subtitle files. For our first release, we set out to tackle and resolve as many of these issues as we could, and to provide an easy and user-friendly interface for operators to review, process, and correct their caption files.
Here is a subset of those functionalities and a short description for each. Some of these items are complex enough that deserve their own dedicated blog. Hopefully soon!
How many times have you been bothered by the fact that the caption of a show is just a tad bit behind or ahead of the actual dialog? The actor stops talking and the caption starts to appear! Or the caption and audio seem to be in sync but as time goes by, there seems to be a bigger and bigger gap between what is being said and the caption that is being shown on the screen. It makes watching a show with closed caption/subtitle quite annoying. There are many reasons for such sync issues, which I will leave for a different blog. But suffice to say, fixing such sync issues is a very time-consuming effort and probably as challenging for the operators who have to deal with them, as it is for you and I who want to watch the show! The operators have to spend painstaking time, adjusting the timing of the closed captions all the way through, making sure that fixing the sync issue in one section doesn’t have a ripple effect of causing sync issues elsewhere. The time to fix the sync issue could vary from a few hours to more than a day!
CapMate, with the use of Machine Learning techniques, can provide a very accurate analysis of such sync issues, determining what type of sync problem exists, and how far off is the caption from the spoken words. And deploying a complex algorithm, CapMate can actually automatically adjust and correct the sync issue throughout the entire file at the operator’s press of one button! This action alone can save a substantial amount of an operator’s time, with amazing accuracy. Users can also perform a detailed review of the captions using CapMate viewer application and perform manual changes.
Another item that can be annoying to an audience, is when the caption text, usually placed at the lower part of the screen, overlaps with burnt-in text in the show. Operators need to manually review the content with the caption turned on to see if and when the caption may overlay a burnt-in text present on the screen. This is another time-consuming process.
CapMate, using a sophisticated algorithm can examine every frame and detect any text that may be part of the content. It can then mark all the time codes where caption text is overlaying on the on-screen text, simplifying the process for the operator who can quickly adjust the location of the caption and remedy the issue.
While this sounds similar to the previous feature, it is actually quite different. There are instances where due to missed caption timing, the beginning of a caption may occur before the end of the previous caption. That, as you can imagine, has a big impact on the viewing experience and is not acceptable.
CapMate can easily detect and report back on all instances where such caption overlaps exist and like many of its other features, CapMate provides an intuitive interface for the operator to have CapMate make the necessary adjustment to all affected captions.
SCC (and other) Standards Conformance:
Closed caption and subtitle files come in many different formats. One of the oldest and most arcane formats (and yet quite prevalent) is called SCC, which stands for “Scenarist Closed Captions.” It’s commonly used with broadcast and web video, as well as DVDs and VHS videos (yes, it is that old!). It has very specific format specifications and is not a human-readable file. Therefore checking for format compliance is a very difficult task for an operator, always requiring additional tools. And making corrections to such files is even more difficult as it is easy to make matter worse by the smallest mistake. There are also a variety of XML-based caption formats that while more human-readable, are still difficult to manually verify and correct.
CapMate has automated Standards conformance capability, and can quickly and easily not only detect file conformance issues for SCC and other formats, but it also can make corrections accurately, and effortlessly. There are a variety of different templates defined for IMSC, DFXP, SMPTE-TT, etc, which CapMate can verify for conformance.
Profanity and Spell Check/Correction:
While some content may include profanity that is spoken, many broadcasters may choose not to have such words spelled out as part of the caption/subtitle. In many cases where automated speech-to-text utilities are used to create the initial caption files, such profane words are transcribed without any discretion. And in case of human authoring where captions are generated manually, spelling mistakes can be easily introduced by the authoring operators.
CapMate provides quick and accurate analysis of the caption text against a user-defined profanity database, and a user-extendable English dictionary to detect both profanity and spelling mistakes. Similar to word-processing software, CapMate allows the operator to do a global replacement of a profane word with a suitable substitute, or fix a spelling mistake. This work will take a fraction of the time using CapMate compare to manual caption/subtitle detection and correction.
Many other Features:
To detail all the features of CapMate here would make this a very long blog! Suffice to say, there is a wealth of other features that deal with items such as CPL (Characters Per Line), CPS (Characters Per Second), WPM (Words Per Minute), or number of lines, that CapMate can verify and provide an intuitive interface for the operator to fix.
I will have to leave those for a separate blog (it is called job security! J )
But if you want to get more details about CapMate please go here or contact us for a demo and free trial! You can also check out the launch video here we made announcing CapMate!