Iced Tea Pour Over Ice
Iced tea over ice in morning window light, one pull-back that keeps the label legible throughout.
Fills the prompt and every setting below into the Seedance 2.5 form. Nothing is submitted until you press generate.
The prompt
Amber iced tea pours from the jug over a tall stack of ice, filling the glass while the label on it stays sharp and unaltered. The camera dollies back slowly and evenly, never losing the label. Bright morning window light enters from the left, refracting through the liquid into warm streaks across a linen tablecloth. Condensation gathers and runs in thin lines down the outside of the glass. Audio: a steady pour, ice shifting, then a soft settling fizz.
Settings this prompt was written for
| Field | Value |
|---|---|
| Mode | Image to video |
| Duration | 5 seconds |
| Resolution | 480p |
| Aspect ratio | adaptive |
| Audio | Audio on |
| Credits | 87 |
They are the real price of that exact configuration — computed with the same pricing engine the API charges with, at the resolution, duration and audio setting shown on the card. Change any of those in the generator and the price updates with it.
Why it works
A pull-back is safer than a push-in for a pour, because the stream stays in frame the whole time while the label shrinks rather than disappearing behind the jug. The light is placed on the left so the refracted streaks land on the tablecloth and give the glass an edge, and condensation is written as a secondary action so the shot has something alive in frame after the glass fills. Audio is a three-part bed — pour, ice, fizz — because that order is what makes a drink sound cold.
What to put in the reference image
A photo with both the glass and the jug visible, the label facing the lens and the tabletop simple. Avoid photos that are already full of liquid motion or cluttered with props; the model only adds the pour, so a clean upright glass gives the most predictable result.
More prompts like this
Prompts stay in English on every language of this site: that is the language the model was trained on, and translating a prompt changes what it renders. The page around it is translated.