M+E Daily

By Ken Kiers, Executive VP Post Production Services, and Mel Waddell, VP Post Production Services, My Eye Media

Welcome to the age of machines. The thermostat can learn your habits and adjust accordingly. Cars can drive and park themselves. Computers can now perform quality control on your content! It may seem as though society is well on its way to robot Armageddon, but it is possible to harness this world of automated QC for the greater good of all humankind.

The feature sets of most automated QC platforms are extensive. It is tempting to turn on all the features and let it rip; however, with each additional parameter tested, the processing burden intensifies — tying up resources and requiring more time. It then becomes necessary to review every flagged issue and filter the true errors from non-issues. The greatest advantage of adding automated QC to a particular file-based workflow is efficiency. The key to unlocking its benefits is to find what the machine does well…

Harnessing the machine

Specification checks are where automated QC excels. With the increase in digital distribution avenues, there are a greater number of similar but different mezzanine files for QC operators to navigate.

In the automated QC template, operators can set up simple checks to verify codecs like ProRes, MPEG2 and J2K as well as corresponding bit rates. At the same time, the appropriate containers like .mov, mxf, .avi, .ts, .ps can be verified. With this process, critical-fail assets can be flagged before wasting any operator time. Requesting replacement assets will make the flow of other jobs continue with minimal effect on production deadlines. During a full linear pass, QC operators can add luminance and resolution checks to the video template that can help them review flagged areas. Another major benefit of an automated QC workflow is the ability to run a photosensitive epilepsy (Harding) test.

This is a specific test for flickers, patterns and luminance shifts over a period of frames that could induce seizures in people who are susceptible to this type of stimulus. More and more territories are beginning to require this test for compliance, which makes automated QC an extremely useful tool. Audio checks and analysis layered with multiple configurations, channels and streams can be time-consuming and confusing tasks for operators. For instance, the typical eight-channel (5.1 + 2.0) audio configuration presents a lot of possibilities.

Within the .mov container this can be configured as eight discrete audio channels. It can also be configured as two streams: one stream with six discrete channels and the second stream with two channels interleaved.

Similarly, it can technically be four streams with two channels interleaved in each one. In all possible cases, the audio will look correct on a scope of a play-back device. However, issues can arise down-stream if the packaging is not properly identified prior to entering the transcoding pipeline. Additionally, audio peaks and averages can be checked based on a client’s specs as well as compliance with loudness regulations.

A file-based workflow analyzing audio for ATSC 85 or EBU-R128 is faster than real-time, which is invaluable when processing a large number of files.

Each use case will be different depending on what stage of the process operators are currently working on. Integrating automated QC into high-level mezzanine file evaluations can provide a comprehensive set of spec checks that will improve the overall product. Even with these advancements in technology, there are still numerous human-centric qualities that are unmatched by automated QC.

Advantages of human touch

While automated QC is extremely effective in providing a preliminary analysis of file metadata and technical specifications, it can-not always accurately interpret some of the most complex data. Additionally, automated QC cannot investigate beyond the initial flagging of anomalies. Although it would appear to be more time and cost efficient to eliminate human QC altogether, its auto-mated counterpart simply cannot replace a human being’s knowledge and experience.

One example where automated QC falls short is in accurately assessing audio sync. Even the most advanced automated QC system can-not verify audio sync down to the frame. Such systems also cannot accurately determine when audio sync drifts from scene to scene or, in some cases, shot to shot, as is often the case.

A skilled human QC operator is capable of not only recognizing asynchronous audio within a single frame tolerance, but also providing individual and detailed examples, including a scene description, line of dialog, or both. Automated QC also relies on the accuracy of the metadata track labeling at the time of the file creation for audio configuration analysis.

Automated QC does not inherently distinguish between English or foreign audio, nor is it capable of deciphering dialogue from music and effects tracks. An automated QC system cannot actually listen to a file to know the dialogue is French instead of English. Automated QC will also not flag something such as dialog leak on an M&E-only track. Human QC still plays a key role in text sync and accuracy verification. Automated QC can verify that closed captioning is present in a file, but not text that is burned into picture.

Furthermore, automated QC cannot yet analyze whether text is correctly positioned or timed with picture and audio. Spelling and grammatical checks are also outside the realm of automated QC capability. Plot-pertinent building signs, newspaper headlines, graphics or other creative choices that include misspelled or foreign language text cannot be validated in automated QC either. Human QC operators also often rely on prior creative notes from clients to assess text accuracy. This type of information cannot be fed into an automated QC program.

As a result, creative content analysis can-not yet be integrated into the automated QC process. The integral part of human QC is verifying that the content is as it is intended to be. Unlike automated QC, these are questions operators can accurately answer: Is this the correct episode or version of a feature? Is it the correct aspect ratio? Subtle color correction and framing discrepancies require the finely trained eye of an experienced human operator. Similarly, automated QC cannot detect visible production crew and equipment.

It is true that automated QC can be programmed to find many anomalies. Systems can be set up to analyze for digital hits, blanking er-rors, black frames and freeze frames.

However, this analysis relies on setting a series of parameters and tolerances to flag changes in picture and audio. No matter how much time is spent fine-tuning these parameters, even the most advanced systems still miss certain anomalies and/or misinterpret creative elements as errors.

Thus, any automated QC report still needs to be reviewed by a human operator for both accuracy and to subjectively rate anything flagged.

Balance is key

Without debate, technology has aided in the exponential growth of human productivity. In the case of QC, finding the balance between technologies, human perception and judgment is the key to a more efficient workflow and provides the most thorough and best results.

Click here to read the .pdf

——————————————-
Ken Kiers has over 20 years of post production expertise. He leads My Eye Media’s development of new digital workflows, including the integration of 4K and HDR technologies. Mel Waddell has 10 years of experience in post production, focusing on quality control and digital workflows.