ISDCF first introduced its Plugfest testing with the goal of evaluating support for, and interoperability of, SMPTE DCP distributions. One of the weak spots quickly found was that of subtitles. The problems were several, as we’ll briefly review. While there have been substantial improvements, not all problems have gone away. These deserve a special look.
Once upon a time, all projectors in the field were based on Series 1 technology from Texas Instruments, called CineCanvas™. The Series 1 projector had a text and graphics rendering engine built-in. In addition, TI specified the nature of the text file for carrying the text and graphics information in the distribution. The result was that one company was responsible for both file specification and rendering, and everything worked fine. Or at least, everything appeared to work fine.
The introduction of the Sony projector, coupled with the introduction of servers for TI projector that rendered subtitles directly in the media block, opened the door to interoperability issues. There are several places where things can go wrong. Text font, location on screen, and synchronization with the image are primary areas. A subtitle/caption designer who makes decisions using one projector may find that their choices look very different when other rendering engines are used. Interestingly, no tools were introduced for helping manufacturers improve interoperability.
SMPTE replaced the proprietary TI file with a new XML-based file, with SMPTE Standard ST428-7. This is the subtitle/caption file to be used for both open and closed text presentation in cinemas. However, as it’s only included in the SMPTE DCP, it is not yet in use. It was the interoperability of this file that first concerned Plugfest testing. In the course of these tests, it was quickly learned that the problems of common font size and position on screen were significant issues, and needed much attention. There were other problems, too, more specific to the variety of ways in which subtitles could be prepared.
This became cause for further testing, and the introduction of test materials, in particular the “Fox Killer Reel,” and more recently the “Fox Sync Test.” Both of these test files have become part and parcel of ISDCF Plugfest testing. The Killer Reel examines rendering issues when faced with every possible subtitle/caption event that can be programmed in the standard distribution file – including font and position. The Sync Test looks for synchronization issues between subtitles burned into picture, rendered on-screen for open subtitles and captions, and rendered off-screen for closed caption applications.
Surprisingly, the Sync Test is revealing more errors than expected. In several of its Plugfests, ISDCF members have noted “rubbery” subtitles. This received more exploration in the most recent round of tests. When these tests first began, it was the closed caption application that appeared to be losing synchronization. To better examine the problem, subtitles were also rendered on-screen for comparison. The results were inconclusive. It wasn’t possible to tell where the errors really were coming from. To better test synchronization, “burned in” subtitles were also generated, such that the subtitle was part of the picture. The timing of rendered subtitles could then be compared with those that were burned in. The results from these tests have illuminated a problem area that appears to have been around awhile: rendered subtitles are not always in synchronization.
Due to time constraints, ISDCF didn’t run repeated tests to determine if the problem was one of erratic synchronization, or if it was more systemic. Since then, other labs have chimed in with results of their own. It appears that text rendered on Series 1 projectors and at least by one stand-alone server is within a frame of where it should be. Surprisingly, Series 2 projectors appear to generate “rubbery” renderings. As the topic receives more attention, more comments are coming out of the closet. In this month’s ISDCF meeting, a major mastering house reported that synchronization of rendered subtitles is “unreliable.” This was a stunning statement, coming at a time when such a basic feature was considered to be mature.
There is more to be learned on this subject, and we can bet that several organizations are now performing deeper examination of the case of the “rubbery” subtitles. No doubt they will stay busy for awhile. If rendered subtitles are having difficulty with 24fps content, just wait until 48fps per eye content is subtitled.