What to do about YouTube auto-generated closed captions?

  • 1
  • Question
  • Updated 7 months ago
  • (Edited)
In the process of working with faculty to develop online (and F2F) courses in Canvas, I often see the use of YouTube videos that include auto-captioning.  While the captions are often fairly accurate, they do contain some mistakes, as well as a lack of punctuation, capitalization, and ambient sound effects.  I realize this type of captioning is not ideal, and it sounds like it does not meet the legal requirements for accessibility.  Yet this is the majority of video content for many faculty. 

Should I tell them this content is simply inaccessible, does not meet legal requirements, and must be removed?

The typical responses to this question are unrealistic:

"Contact the source of the video and ask them to have it captioned" (good luck)
"Find another version of the video that is correctly captioned." (often impossible)
"The instructor can caption it themselves using Amara." (are you kidding?)
"Download the video and caption it yourself." (generally illegal, for one thing)

While I am highly sympathetic to the need for accessibility, I also have a hard time telling faculty that they need to remove so much of the content they have.  This is often crucial material for the class, and not easily replaced.

So, what's the final word on auto-captioned YouTube videos...can they be used, or not?

Thank you.

Photo of Chris P.

Chris P.

  • 8 Posts
  • 0 Reply Likes
  • in a grey area.

Posted 8 months ago

  • 1
Photo of Sean Keegan

Sean Keegan, Official Rep

  • 20 Posts
  • 9 Reply Likes
Official Response
Hi Chris,

In the majority of cases, the use of auto-captions for YouTube videos will not be appropriate. While the auto-caption functionality has improved significantly over the past 8 years, it still can mis-recognize words, phrases, and punctuation to the extent that the caption information is not an accurate representation of the audio track.

Additionally, if there are multiple speakers, the auto-caption functionality does not specifically distinguish or identify who the speakers are nor will it regularly identify sound effects that may be necessary. Given this information (and the aforementioned issues), this can lead to issues with understanding the information the the video. From this, I would say that YouTube auto-captions are not sufficient.

That said, I do believe that YouTube auto-captions can be an excellent starting point for editing or cleaning up a caption file. The accuracy is far better than in previous years and, often times, the editing required is minimal; rather than rewriting entire sentences, it is often a matter of fixing proper nouns or inserting punctuation.

I have seen *a few* videos in which there was only one speaker (who spoke clearly and at an even pace) in which the auto-caption functionality was highly accurate and *might* be acceptable. But, even then, there was still a need to make a few corrections, so, technically, it could be argued that the auto-caption functionality still did not meet the definition of "captions".

So, short version is that I would not accept YouTube auto-captions as being sufficient at this time. They are indeed a very good starting point, but in many cases do still require some editing to be considered an accurate representation of the audio track.

In terms of what to communicate to faculty - I think it is important to inform faculty of what the issues are and where there are problems/gaps in the current technology. It may be part of a larger conversation that requires discussion with department chairs (and/or deans) as to how to best support faculty who use audio/video content as part of any instructional materials for a course. Often the focus is on how faculty will caption and make these materials accessible, and I agree with you that this is not necessarily a realistic solution. At the CCC Accessibility Center, we are working on a few captioning related projects, but none have necessary combination of technology to be successful at this time.

If there is a significant amount of video content be utilized for online instruction, then rather than focusing on individual faculty to address the issue, a more cohesive campus-wide solution may be more appropriate. What could be explored is what resources may be available from a campus perspective to support the editing and revision of such captions. Some colleges use student workers to help develop captions whereas others have decided to outsource to third-party vendors. This does not solve the immediate problem, but may help start a conversation as to how to develop a more functional solution across the institution.

While there is some support in the CCC system for captioning instructional materials (e.g., the DECT grant), I do recognize one obstacle is that this funding resource is not available to caption YouTube content. Addressing that issue would require discussion with that grant administrators and resolved through separate channels.

To summarize - No to YouTube auto-captions and yes to a different approach at the campus level.

Hope this helps. Please feel free to ask any questions or identify concerns.

Take care,