● Gemini Omni Prompts
Home / All Prompts / avatar

AI Avatar Summon: @username + Background Change in One Prompt

Use Gemini Omni Flash's @username avatar syntax to summon your digital likeness into any scene — verified working pattern from Chrome Unboxed's hands-on review.

avatar 16:9 10s Conversational summon #avatar#omni-flash-exclusive#username-summon#talking-head

Prompt

Create a video of @your_google_username speaking warmly into the camera.
Background: a modern penthouse view of the Chicago skyline at golden hour,
visible through a wall of glass behind them. Keep their face, hairstyle,
and clothing consistent with the saved avatar. 10 seconds, chest-up framing,
natural lip-sync, slight head movement for liveliness.

Why this prompt

This is literally adapted from Chrome Unboxed’s hands-on review of Gemini Omni Flash’s avatar feature, which is a Flash-exclusive capability (not in Sora, Veo, or Runway).

The Chrome Unboxed editor’s actual test prompt was:

“Tell me about how easy it is to make a Gemini Avatar. Also, make my background a penthouse view of the Chicago skyline.”

Editor’s reaction: “facial tracking, micro-expressions, and lip-syncing are shockingly good”.

Source tier: 🟢 Official feature + first-party media hands-on (high confidence)

The @username syntax (Gemini-specific)

Per Google’s official Avatar help page, once you record your avatar reference once, you summon it in any future prompt with @your_google_username (use your actual Google account name).

Official examples from Google Help:

  • Create a video of @[your Google username] singing with an orchestra.
  • Write a story where @[your Google username] is the main character, facing a dragon.

Hard limits (Google’s own guardrails)

Critically important — most third-party prompt sites omit these:

  • Age: must be 18+
  • Geography: Avatar feature not available in EEA / Switzerland / UK
  • Language: English-only at launch (May 2026)
  • Watermark: every video carries Google SynthID — non-optional, embedded in pixels
  • Reference recording requirements: clear eyes/nose/mouth, no sunglasses / masks / hats covering face, no other faces in the background, well-lit

How to tweak

  • Stay in lane: avatar works best for talking-head scenarios. Complex hand gestures and full-body motion are unreliable (industry-wide AI video limitation)
  • Background swap is the easy win: avatar in cafe / studio / outdoors / fantasy scene all work as long as the avatar is foregrounded
  • Single-language pivot: for multilingual videos, regenerate per language using the same @username reference — Omni preserves likeness across languages

Common failure modes

Per digit.in’s review and Chrome Unboxed’s caveats:

  • Complex hand articulation (counting fingers, sign language, instrument playing) — drifts
  • Multi-person scenes — Omni refuses if other faces are in the reference
  • Crowd backgrounds with visible faces — also degrades

Notes

  • Avatar is Gemini Omni Flash exclusive (not in standard Gemini)
  • SynthID watermark is invisible to humans but detectable by Google’s tools — disclose if using for advertising

Sources