How Sylfayra's multilingual talking avatar works, and when to use it

The avatar mode takes a single reference photo and a text script, and generates a video of that person speaking the script with lip-synced audio — currently built on Veo 3.1 Fast under the hood, so it inherits its 1080p quality and native audio.

It's a genuinely different use case from our other text-to-video models: you're not describing a scene, you're producing a talking-head video of a specific person (or yourself) saying specific words.

Where it works best: video messages to family in another language, quick social clips, presenter-style intros for a product or service, or trying out a script in multiple languages before recording it for real.

Where it's not the right tool: full scenes with camera movement or multiple people — for that, use one of the text-to-video models instead and describe the scene directly.

One practical tip: since this mode is billed like any other generation, write out your full script first and proofread it before generating — a shorter, well-edited script generally reads more naturally than a long one.