Many people ask if edits in the transcript get reflected in the subtitles and vice versa. Here is why this is usually not the case.

The process of creating subtitles usually starts by transcribing the audio. When using third party software, transcribing can be done directly in the subtitle editor. However, we recommend to polish the transcription using the transcript editor prior to generating the subtitles. It gives you a better overview of the content and it saves time.

In either case, as a subtitle editor, you don't want the subtitles to be a literal representation of the words in the audio for a number of reasons:

  • For most types of content, the reading speed of a typical user (preferably lower than 10 characters per second) is 50% slower than the recommended speed of a news anchor or voice over (up to 180 words per minute). 
  • Subtitles should not contain repeated words or redundant content, as is often the case in natural spoken language.
  • A subtitle operator strives to draw as little attention as possible from the main content.

For such reasons, subtitles usually contain less words than the audio. For example, a subtitle operator would replace the literal "Hi Nancy, what is your exact location?" by the better "Where are you?"

So given that the transcription of the audio and the subtitles associated with the content are not linked one-to-one, here is how we recommend to produce subtitles when starting from the audio transcript.

First of all, in cases where the audio is good enough to allow machine transcription, we recommend to use automatic transcription. Depending on the quality of the sound and the speech, the accuracy will be somewhere between 70% and 95%. One hour of content takes about 10 minutes to transcribe, so this step saves you a lot of time.

Next, we advise to polish the transcription prior to cutting it in subtitles, as it is more convenient to correct the words in a text editor than in a subtitle interface. Polishing the transcription includes correcting the segmentation according to the different speakers, correcting words, or adding or removing some fragments.

When the transcript is more or less correct, you can instruct us to cut the transcription into closed captions. We'll take into account the applicable spotting rules, which you can configure taking into account your specific timing and styling requirements.

Finally, there is the quality control of the subtitles. As the logic to create the subtitles isn't allowed to alter the content of the transcription, some individual subtitles might be too long or too short, or the word rate might be too elevated. As a subtitle editor, you can use the interface to modify the contents and or the timing of individual subtitles to make sure the subtitles are properly timed and styled.