Interra Systems localisation3

Netflix now streams in 190 different countries, which means that video services are no longer specific to a certain region. Their reach is global. Content globalisation has prompted many providers to find new ways to speed up video delivery while maintaining exceptional quality.

Manik Gupta is Associate Director of Engineering at Interra Systems and told Digital Mdia World that although creating content for different locations is an opportunity to reach new audiences, preparing content in various languages can be a challenge. “Roughly 6,500 different languages are spoken around the world today, each with its own distinct characteristics,” he said.

Language-Specific Content

“For years, content providers relied on captioning and subtitles to deliver content in regional languages. However, subtitles and captions have certain limitations. Subtitles may require compromises to deliver long dialogue scenes within limited screen times. They can create a distraction and often fail to convey the emotions in a dialogue.”

Audio dubbing is an alternative that gives content providers an opportunity to add language-specific content to the original audio. For broadcasters and content providers today, the challenges of audio dubbing, and the need to automate quality control (QC) in audio dubbing workflows, are more significant now than they were earlier on. Recent developments are helping to improve the efficiency and the quality of audio dubbing.

Interra Systems DMW Audio

Audio dubbing translates a foreign language program into the audience's native language. Given the limitations of subtitles, in many contexts, audio dubbing has become a preferred option for delivering content to many audiences around the world. Therefore, when content creators produce dubbed audio tracks, they must take care to ensure the quality of the recording, as synchronisation issues remain one of the biggest challenges facing dubbing.

Making Dubbing Simpler

Manik said, “Any type of noticeable lead or delay between a piece of audio and video can have a negative effect on the quality of viewing experiences. For example, the time it takes to communicate the same message in a different language, for example Arabic vs Hebrew, can vary drastically, creating synchronisation problems. Sync issues also occur when a frame on the dubbed track is lost.

“Therefore, content distributors need QC methods that are efficient at checking synchronisation between dubbed tracks and the master tracks, and capable of determining if an audio track has changed in any way after the dubbing stage. Any mismatch in the timing between original and dubbed material will cause synchronisation problems that lead to viewer dissatisfaction.”

Taking into account how widely distributed video content is today, providers are handling massive volumes of content, making it essential to deliver a high quality of experience to consumers. “Viewers today have a low tolerance for poor video quality, and audio/video sync issues can be among the worst offenders in terms of causing viewer dissatisfaction,” said Manik.

“By automating QC workflows and performing comprehensive checks, content providers can make audio dubbing simpler. Advances in automated and AI-/ML-based systems are improving QC workflow efficiency, allowing content providers to detect audio dubbing issues, rapidly and accurately, with less manual supervision. In five ways particularly, such systems are completely changing audio dubbing QC for content providers.”

Interra Systems localisation3

Automating Viewer Satisfaction

When verifying dubbing content with complex structures, automated QC systems are making the process more efficient, including multiple MXF and .wav files, to make sure that content variations are accurate and that audio tracks are dubbed properly.

By using an automated QC system, content providers can certify that metadata properties are precise, and also check that the number of audio tracks, channel configuration of dubbed tracks, and duration of the original audio track compared with dubbed audio tracks are correct.

Automation simplifies synchronization checks in a few different ways. First, content providers can rapidly confirm synchronization between video and dubbed tracks, as well as between original and dubbed tracks.

An automated QC system can detect whether synchronization issues exist between the video and dubbed tracks by verifying the presence of black frames in video tracks when there is silence, as well as by detecting colour and test tones in dubbed audio tracks.

However, checking for synchronization loss between the original and dubbed audio tracks is somewhat trickier, since the audio data are going to be vastly different. Even so, the two tracks have common background music or effects, which can be separated from the audio track using techniques such as bandpass filter, which passes frequencies within a certain range and rejects those falling outside that range.

Automated QC that can check for localised correlation between background beds of dubbed audio tracks and the original audio track, and compare loudness curves, will ultimately help content providers identify audio that is out of sync.

Globalised Video

Language detection and identification is an important part of the audio dubbing process. “Over the last few years, AI/ML algorithms have become so intelligent that automated QC systems can detect language in any audio track with more than 90% accuracy,” Manik said. “It only takes a few hours to train AI/ML models before they’re capable of predicting the dialect spoken in the audio track. Using metadata, content creators can verify that the language detected in the audio track is correct.”

Interra Manik Gupta

Manik Gupta, Associate Director of Engineering at Interra Systems

Many countries today have regulations requiring that regular television programming and commercials have the same level of loudness, put in place to ensure that commercials are not substantially louder. With an automated QC system, content creators can compare loudness values between audio tracks based on standard CALM or ITU algorithms.

“Thanks to audio dubbing, content creators can globalize their video and reach new audiences in different areas of the world,” said Manik. “Automating QC dubbing workflows has become critical to helping content providers speed up what has been a time-consuming, painstaking, manual process. With automated QC and the use of AI/ML systems, dubbed content can be produced with greater speed and accuracy, ensuring a high quality of experience for global viewers every time, whatever device they choose to view content on.”