The first step for either of the above options is to get an exact transcription of the video. This is time consuming and error-prone for people who are not transcriptionists. In addition to accurately capturing the dialog (which for impromptu speech is often disjointed), the transcription should identify who is speaking, when there are multiple voices, and describe sounds. From our experience it is best done by editing a file created with speech recognition or by paying for a professional transcription. The transcription process is often combined with the process of creating the caption file, which includes timing information. If captioning is needed to meet the needs of your students, Student Disability Services at Cornell will have additional options. 

If you will be creating an open captioned video, the transcribed text is added to the video using features of the video editing software you are using. Some programs, like Adobe Premiere, will use automated speech recognition to capture and place the text. 

For closed-captioned videos, timing information may need to be added (if it was not done in the transcription process) to create the caption file. 

There are several formats for timed-text files. The Kaltura system used in Canvas requires .SRT or DFXP files. If you are hosting your own video on a website, the different video formats can require other formats. All Center for Teaching Innovation video-related services (Kaltura in Canvas and Panopto) support caption files. Upon request, it is also possible to use a section 508 compliant video player to playback videos in your course or on your web site.