Avatar script CTA: no click the green button, WCAG 1.3.3

The rule and the fix

Name the control, not its look or position. The Understanding page for 1.3.3 (read 2026-10-04) is Level A and says instructions must not rely solely on sensory characteristics such as shape, color, size, visual location, orientation or sound. Its example failure is an instruction that tells people to look for a blue diamond; a passing instruction uses a label such as Next along with any positional cue.

A spoken avatar call to action is an instruction. Saying press the green button on the right relies on color and location alone; saying select Start free trial names the control.

Why avatar scripts are at risk

A script for a Sume talking video is written once and rendered into a 4 to 60 second clip, so its wording is frozen into pixels and audio. A page redesign that moves or recolors the button leaves the old words in the MP4. A label-based script survives the redesign if the label stays; a color-and-position script does not.

Rewrite patterns

Swap the sensory cue for the label.

A short checklist

Run it before you submit the job.

• Does every instruction name a control by its label?
• If a color or place is mentioned, is the label said as well?
• Is the label on the page identical to the one in the script?
• Will a re-render be needed if the label changes? Plan for it.

Keep copy and page in sync

Because the clip is rendered, treat the script as part of the page's source. When you change a button label, re-render the clip. For a series of steps, one job per step, as in the onboarding checklist post, keeps the fix to one short clip. Publish the script as text too, per the text alternative post, so the instruction is available to everyone in the same words.