A Video and Media Review Asks Whether Everyone Can Press Play

What We Keep Finding

Try this with your best-performing video. Play it on mute and see how much of the message survives. That is roughly the version a deaf viewer gets, and if you are leaning on the platform's auto-captions the survival rate is worse than you think. Auto-captions are a rough draft, and W3C's own worked example of what goes wrong is not a mangled product name. It is a dropped negation. Spoken "you should not preheat the oven", captioned "you should know to preheat the oven". It reads perfectly. It says the opposite thing.

Accuracy here has a definition, which helps, because most teams assume it means word-for-word and give up. It does not. Captions have to carry the dialogue verbatim or in essence, plus the sounds that matter, and simplifying is expressly allowed. That is a bar a human editor can actually clear in an afternoon per video, and it is the bar we review yours against.

Captions are only half the assignment, though, and the cheaper half of the other one surprises people. A blind visitor needs the player itself to work by ear, so buttons that announce as more than "button", a seek bar the keyboard can move, and audio description wherever the visuals are doing the talking. Audio description at the level most sites are held to has no substitute. But one rung lower, a full transcript satisfies the requirement on its own, which means a written transcript is a real conforming answer and not a consolation prize. For a library of old videos that is the difference between a project and a to-do.

One last thing about autoplay, because it gets filed wrong constantly and the wrong filing costs you an option. Video that moves and makes no sound has to be pausable. Audio that plays by itself for more than three seconds has a second route out. A control that changes its volume independently of the system volume satisfies that rule just as well as a pause button does.