Wine bottle reveal clip: 6 seconds of Wan 3.0 from a label photo

To animate a bottle into a 6-second reveal, send your label photo as the first frame to wan-3.0 and prompt only the camera and light. It costs $0.38 at 480p, $0.75 at 720p and $1.50 at 1080p on Sume. Pick 1080p for a product page, because the label is what a buyer reads.

The risk is the label, not the glass. Generated video can warp small type, so the first thing to check on every take is whether the label text survives the whole six seconds.

Price

Wan 3.0 lists at $0.05, $0.10 and $0.20 per second and Sume bills list times 1.25, rounded up to the cent (video router docs).

Prompt for a reveal

Keep the motion simple and the subject still: a slow push-in while warm light sweeps left to right across the glass. Do not ask for pouring, because liquids and hands are where artifacts show up. Use frame_images with frame_type: first_frame for the photo (video generation docs).

• Photo with the label flat to camera, even lighting, no glare.
• One verb in the prompt: push in, orbit, or rack focus.
• Duration 6, resolution 1080p, generate_audio false if you will add music.

A last frame for a clean ending

wan-3.0 also supports a last frame. Send a second photo, such as the bottle beside a glass, and the clip travels between two real images. That keeps both ends of the clip faithful and leaves the model to invent only the in-between.

Compliance

Alcohol advertising has rules that vary by country and platform. Sume does not check them. Have a person review each clip and its caption before it runs.

Practical notes

Lighting is the second thing to control. Shoot the label photo against a plain background with soft light, because reflections on glass can turn into odd shapes once the camera starts to move. A matte backdrop gives the model less to misread.

If the label text does not survive, do not re-roll endlessly. Keep the best take, remove the label region in your editor, and place the real label artwork over it as a still or an overlay. That costs nothing extra and keeps the text exact.