AI and closed captioning

The need for closed captioning continues to grow as the way we consume content has increasingly moved online. Even though closed captioning is usually associated with making content accessible for hearing-impaired viewers, it also allows a wider audience to better understand what is going on in the content whilst watching it in noisy spaces like restaurants or with the volume turned down in sound-sensitive environments.

The cost for manual production of closed captioning is expensive for some, so it is clear why a lot of companies would favour AI over humans. However, professional closed captioning transcribers are typically highly accurate – usually around 99% – and it is hard for a machine to match this.

Some of the most common causes of errors when using AI include:

  • Multiple speakers or overlapping speech
  • Background noise or poor audio quality
  • False starts
  • Acoustic error
  • Punctuation, grammar, and numbers
  • Non-speech elements

Having said this, even with a video of perfect quality, there still tend to be errors when using AI. We recently completed a project to test the most up-to-date technology available with a perfect-quality video and as you can see from the screenshots below they aren’t 100% accurate and tend to not make sense.

  • Facebook
  • Twitter
  • Pinterest
  • LinkedIn
Let me draw your attention to the word carrot!

  • Facebook
  • Twitter
  • Pinterest
  • LinkedIn
I don’t even know what’s going on here! A good example of losing complete sense of what is going on in the video – which is of no use, especially to a hearing-impaired audience.

We Still Need Humans

There is currently no fix for the challenges we face with AI closed captioning so for now it will not come close to the accuracy of professional captioners. At McGowan Transcriptions we can offer two solutions for closed captioning depending on your needs and budget.

Firstly, we can offer you our professional closed captioning service. This is where a human will sit and watch the video whilst transcribing what is being said on a separate file, which is later synched to the video using time stamps. They will go over it to ensure the accuracy of the closed captioning is 99.9%, if not more! This is obviously the more expensive option of the two but you will know your job is in very safe hands and you will not need to edit the version you are provided. The captions will be correct and make sense – which is essential for those who rely on captions.

However, we understand that humans completing the process is likely to be too expensive for some companies, which is where our second option comes in. We are currently in the process of creating a bespoke AI system which will automatically generate captioning for your videos. With this option though, you don’t get the guarantee of accuracy so you would need to go through this once the AI has generated the captions to amend anything and ensure it makes sense. This is a good solution for those who do not require the high level of accuracy in their captions and are on a tight budget.

If you have any questions on closed captioning or any transcription work please do not hesitate to contact us!

  • Facebook
  • Twitter
  • Pinterest
  • LinkedIn
Fairly certain these aren’t the correct lyrics… oops!