● Gemini Omni Prompts
Home / All Prompts / avatar

AI Avatar: Talking Head for Corporate Explainer

Build a consistent digital likeness avatar using Gemini Omni's avatar tool — corporate explainer style talking head with natural gestures.

avatar 16:9 30s Corporate explainer #avatar#talking-head#corporate#explainer#ugc

Prompt

Generate a talking-head video using my saved avatar (uploaded reference). The avatar
sits centered, framed from chest up, looking directly into the camera. Background:
modern minimalist office, slightly out of focus, with a single accent plant on the
left edge. Soft three-point lighting, even on the face. Avatar speaks the following
script naturally with slight head nods, hand gestures appearing at the bottom of
frame every 8-12 seconds:

"[YOUR SCRIPT HERE — keep paragraphs under 25 words each for natural pacing]"

Lipsync accurate, micro-expressions natural, no awkward blinking patterns. Maintain
consistent skin tone, hairstyle, and clothing throughout. 1080p, 24fps cinematic.

What this prompt does

This prompt activates Gemini Omni Flash’s dedicated avatar tool, which lets creators build a digital likeness and reuse it across videos without re-uploading reference material each time. The key trick is keeping the avatar consistent — Gemini Omni stores your appearance, so subsequent generations match.

Best use case

  • LinkedIn / YouTube thought leadership videos (when you don’t want to film yourself)
  • Course / online training videos at scale
  • Localized marketing (same avatar, different language scripts)
  • Internal company comms

How to tweak

  • Background: “modern office” → “home library” / “cafe corner” / “outdoor terrace”
  • Wardrobe direction: “wearing a navy blazer, white shirt, no tie” (Gemini Omni respects this)
  • Tone of delivery: add “warm and conversational” or “authoritative and slow”
  • Script density: keep paragraphs short (< 25 words) for natural pacing; long sentences cause robotic delivery

Notes

  • The first generation locks in your avatar — subsequent prompts in the same Gemini app session inherit it.
  • SynthID watermark is non-optional and embedded in every output. This is fine for most use cases but disclose to viewers if using for ads.
  • For multi-language videos, generate once per language with the same avatar reference — Gemini Omni handles localization while preserving identity.
  • 30s is roughly 75-85 spoken words at natural pace.