Internal training and corporate change projects often generate a large amount of content that needs to be made available quickly to employees in another language. However, it is not always possible to wait weeks for classic video processing, transcriptions, translations, and subtitles.
In one project we addressed just such a situation. The customer needed to prepare Czech subtitles for extensive online training videos as quickly as possible and at the same time translate some related PowerPoint presentations. The materials were used for internal employee training when implementing a new company system.
The entire project included approximately 8 hours of video recordings from Zoom calls and extensive presentations that accompanied them. The challenge was not only the volume itself, but also the quality of the original recordings, multiple speakers, different accents, and a very short delivery deadline.
The challenge: speed, volume and clarity
It’s not enough to merely "translate the text” for an assignment like this. First, a reliable transcript of the spoken content is needed, and this must then be timed correctly, translated, and checked against the original recording.
Moreover, the input materials were not studio-prepared. We're talking about recordings of online meetings, where poor sound, overlapping speakers, different accents, and incomprehensible utterances are commonplace.
The customer also needed the output very quickly so that the materials could be used for internal training. It was therefore necessary to design a procedure that would combine the speed of technological processing that was subsequently checked by a human.
Solution: Modern technology and a Human-in-the-Loop
We chose a combined approach for the project. We used modern technology to accelerate processing, followed by human review at every stage.
First, we prepared an automatic transcription of the spoken word and the timing of the individual sections. This was then translated into Czech. Next, a proofreader joined the process and checked the subtitles by listening to them against the original video.
This step was crucial. With automatically generated subtitles, it's not just about linguistic accuracy. The meaning, clarity, correct capture of technical information, names, numbers, context and any inaccuracies caused by poor sound quality or the speakers' accents are also checked.
This allowed us to benefit from the speed of AI technology while maintaining human control over the outcome.
Output according to purpose of use
This was not a case of classic film subtitles designed for continuous viewing without stopping. The goal was to prepare a transcription of the spoken Czech content that was as clear and complete as possible, this was then used by employees during internal training.
Thus, the subtitles were processed in such a way as to preserve all the information from the original speech. This meant the individual subtitles did not have to be subject to strict truncation and could be more detailed in content, so users could return to individual parts, stop the video, or watch a specific section repeatedly.
We delivered the output as videos with Czech subtitles embedded directly into the video. Depending on the customer's needs, we can also deliver subtitles separately, for example in .srt format.
In the case of, say, a marketing video, e-learning content designed for continuous viewing, or content intended for public distribution, we would choose a different approach. We can also prepare classic subtitles from AI transcription, but these usually require additional linguistic, stylistic and creative adjustments, shortening the text, working with line length and adjusting the reading pace.
Result
The project was completed by the express deadline. The output was 8 hours of video recordings with Czech subtitles prepared for internal employee training as well as some translated PowerPoint presentations.
This project showed that for large-scale video content, the combination of AI technologies, translation, and human review can be a very effective solution. Technology can significantly speed up the first phase of processing, but human expertise remains essential for understanding context, correcting inaccuracies and ensuring the final content is fit for purpose.
AI subtitles are not just about automation. They are primarily about a properly set-up process that corresponds to the purpose of the video, the quality of the input data, the requisite deadline, and how the resulting content will actually be used.
Are you working on a similar project and need to make your videos, podcasts, or training materials available to international colleagues or customers? Contact us for a free consultation and quotation.