Google announced image-based custom avatars for Vids on 2 November 2026. In its 3 November update, Google paused the rollout; Rapid Release and Scheduled Release dates are now to be determined. No resumption date is given. If your eligible account offers the avatar workflow, its appearance still says nothing about whether the voice reads your script correctly.
Specify the reading before generating
Suppose a fictional parts guide says “part 120”. If 120 is an identifier, you may want “part one two zero”, rather than “part one hundred twenty”. Keep the displayed identifier intact and prepare the intended spoken wording separately. Include a date, an abbreviation and a unit in the same short review script.
In Vids, choose the avatar and voice, enter the script and use the available preview. The documented audio-tag menu can adjust pace, pauses, emotion and style. It is not a guarantee that every identifier will be pronounced as intended. Use only controls actually offered by the tool, then listen to each marked item.
Check the generated scene again
Listen after insertion as well as in preview. If you change the script, check whether the scene needs updating; the current help describes an “Avatar outdated” indicator. Compare the final audio with the approved script before exporting.
Check your plan’s generation allowance and supported language before regenerating. Do not treat the broad announcement as unlimited or immediately available access. Tell the audience that the narration is AI-generated. This is a proposed check with invented words; no avatar, voice or audio was generated for this article.