GEMINI OMNI PROMPTS

VFX

16:9 10s Trigger-based VFX

vfx

VFX Trigger Pattern: Mirror Ripple Transformation

Adapted from DeepMind's official Gemini Omni hero demo — use 'when X happens, Y changes' trigger syntax for cinematic VFX.

vfx trigger-pattern conversational-editing official-demo transformation

Prompt

Turn 1 — Base scene:
A young woman in a white linen dress steps in front of an antique full-length mirror
in a sunlit Victorian parlor. She slowly reaches her hand toward the glass. 10 seconds,
locked-off shot at chest height, soft natural light.

Turn 2 — Trigger transformation:
When her fingertips touch the mirror, make the mirror ripple beautifully like liquid,
and the person's arm turns into reflective mirror material extending into the glass.
Keep the parlor, her dress, and the lighting identical.

Why this prompt

This is literally one of Google’s official hero demos for Gemini Omni, shown in both DeepMind’s model announcement and the prompt guide.

The “when X happens, Y changes” pattern is now known as the trigger pattern, and Medium’s Gemini Omni Prompt Playbook documents it as one of the 10 examples that consistently work.

Source tier: 🟢 Official Google demo (highest confidence)

Other official trigger prompts (Google’s viral demos)

From DeepMind’s prompt guide:

  • Make the sculpture out of bubbles
  • The lights of the apartments start turning on in sync with the music
  • Change the ships to be made from white origami paper

All four (above + the mirror ripple) were chosen by Google itself to demonstrate Omni’s strength. Adapting these structures gives the highest success rate.

The trigger pattern in detail

The structure that works:

[Base description (camera locked, subject in position)]
+
When [specific trigger action], [specific transformation].
Keep [list of preserved elements] identical.

Why it works (per Medium Playbook): Omni interprets “when X happens” as a temporal anchor, then concentrates the visual change on the trigger moment instead of drifting the whole scene.

How to tweak

  • Subject swap: woman in parlor → child in attic / dancer on stage / fisherman by lake
  • Trigger action: touch → speak → step / kneel / spin / clap / breathe
  • Transformation: mirror ripple → flames erupt / vines grow / sand pours / pixelate
  • Locked elements: always specify what NOT to change

Common failure modes

  • Empty epic adjectives (“epic / dramatic / amazing”) don’t add specificity. DeepMind explicitly suggests using real-world knowledge references instead
  • VFX across multiple shots: stick to 1 trigger per 10-second clip (Flash limit)
  • Locked-off camera (no movement) works much better for VFX prompts than tracking shots — the visual change is then unmistakable

Why VFX is one of Omni’s strongest categories

Per Atlas Cloud’s review, Omni’s physics understanding (referenced in Google’s announcement) makes trigger-based transformations far more believable than equivalent prompts in Sora or Veo. The mirror ripple demo specifically went viral on X / Twitter because the physics felt grounded.

Sources