A reel should not become a different story when someone turns the sound off. Captions help people who cannot hear the audio and people watching without sound, but decorative text alone may leave most of the explanation missing.
W3C's caption guidance explains that captions convey speech and the non-speech audio needed to understand the content. YouTube also warns that automatic transcription can misrepresent speech and needs checking. Do not assume that a familiar Tamil name or a mixed-language sentence was recognised correctly.
Review meaning, then timing
Use an original practice clip in which you explain how to arrange a shared meeting table. Write the spoken instructions, then watch the video muted.
Make three passes:
Meaning: Does the text include the instruction, condition and correction? “Use this side” needs enough context to identify the side. A mistaken “do” or “don't” can reverse an instruction.
Timing: Does each caption appear with the relevant action and stay long enough to read at a normal pace? Pause where you repeatedly have to rewind.
Placement: Is the text visible against the footage, and does it cover the object or hand movement you need to show? Check the exported version on the phone where you expect people to watch.
If an off-camera knock explains why you turn around, identify it. If background music adds no necessary information, avoid filling the screen with descriptions that distract from the actual instructions.
Ask for one specific review
Send a draft privately to a willing friend and ask them to describe the task without hearing the audio. Their answer can reveal an omission; it is not a certification of accessibility.
The close attention used in editing useful writing matters here too. Caption correction takes time and belongs in the production plan, including when somebody else does it for your team.
Keep the final transcript with the project files. When you revise a spoken instruction, update its caption before exporting the next version.