Subtitles are good sound
Most videos are consumed without sound
With the ever increasing use of social media content, the use of text within videos also increases. According to the news site „mic“, 85% of Facebook videos are consumed without sound. If you want to get through with your message here, subtitles in high quality in the respective language are absolutely necessary. Subtitles have now established themselves as the perfect source of text. File formats such as WebVTT and SRT have become established for data exchange on different platforms. The creation of subtitles can be translated manually or automatically by cloud-based services such as Video.Taxi using editors on the computer. The workflow looks like this:
Have the language of the video translated into text as subtitles
With the help of artificial intelligence and machine-based learning, a computer system recognizes the spoken word and converts it into subtitles. The recognition of the words is relatively simple. The challenge arises when multiple results are assigned and context is included. The better the context, the better the subtitle.
Manual fine tuning
No automated detection works flawlessly. With the help of an online editor, our customers can manually correct errors in order to achieve high quality. The subtitles are saved in a separate file. Changes are therefore possible at any time without having to touch the actual video file.
Auto trimming – use of a transcript or protocol
A text version of the spoken word, e.g. a script for a film or a transcript can also serve as the basis for generating subtitles. The advantage here: the text version is 100% error-free. However, the timecode is missing here. When exactly which word should be displayed as a subtitle? The auto-trimming method is suitable for this. The document is uploaded and adapted with the text recognition version including time codes. What remains is the existing text version with current time codes and thus the finished subtitle.
Bundestag TV as a pilot project
Hundreds of hours from the Bundestag TV program were used to develop and bring the above-mentioned method to market. With subtitled content, the Bundestag fulfills its duty to offer accessible content. The protocol service also regularly produces a minutes version of each plenary session that is accessible to the public. Using the video recordings and the protocols, it was possible to develop the system and train it accordingly.