Coffee product video from a photo: steam and subtle motion

A coffee product video works best when the cup remains believable and the motion gives the viewer one thing to notice. Steam and a gentle camera move can add atmosphere without asking the model to invent a pouring action, moving hand and changing pastry at once.

This guide uses our original coffee-and-croissant example. It is an AI-generated still animated into a video, not footage of a customer's café or proof of a commercial campaign result.

The example and its settings

The published take used Kling 3.0 Pro, image-to-video, five seconds, 16:9 and audio off. Its recorded provider cost was $0.56 on September 8, 2026. This is the cost of that selected generation, not a price promise for future runs or a complete advertisement.

Open the coffee recipe to inspect the saved settings, or start a fresh Kling 3.0 Pro generation from the same image.

As a newly verified alternative, Seedance 2.0 runs 4–15 seconds at 480p or 720p with audio optional; a 4-second 720p generation is 303 credits. Try both on the same photograph and compare.

Prepare the photograph

Choose a frame in which the cup rim, handle and saucer are easy to distinguish. Avoid a bright reflection that merges the handle into the background. Leave space in the direction of the intended camera move so that a small change in framing does not crop the cup.

If you need a price or headline, plan a clear area for it but add the exact lettering in an editor later. Keep a clean copy of the image. Reshooting or recropping a confusing input can be more useful than repeatedly expanding the prompt.

Separate what moves from what stays still

The example uses this direction:

Thin translucent steam rises slowly from the coffee and curls naturally in the warm side light. The camera makes a very gentle forward push toward the cup and croissant. The ceramic cup, handle, saucer and pastry remain rigid and completely stationary. Preserve all food textures. Cozy editorial food commercial, single continuous shot, no cuts, no added objects, no text.

The important distinction is steam versus solid objects. A vague instruction such as “make everything cinematic” does not say whether the cup, camera or light should move.

For a simpler variation, hold the camera still and animate only the steam. That is a prompt idea, not another tested take. Compare one change against the baseline rather than changing framing, light and movement together.

Review the full five seconds

Check the cup handle at the beginning, middle and end. Watch the croissant edges for melting or texture drift. Steam should remain a secondary detail; if it hides the product, reduce its density in the next brief.

Use the frame extractor for close inspection, then replay at normal speed to catch flicker. A clean opening frame is not enough to approve the shot.

Turn the shot into a finished product video

To build a complete video around the accepted steam clip, open Product Video Maker from Photos & Clips and combine it with your own café photos or footage. Upload one to three files, write your opening caption, a product detail and a call to action, then review the three editable scenes in Studio before exporting an MP4. Composing your own photos and clips uses no AI generation credits; adding another AI-generated shot is optional and shows its own credit quote before you generate it. The project saves to My projects, so you can reopen it later to keep editing or download the finished file. This assembles the scenes you choose — it does not estimate ad performance, conversions or viral reach.

Turn the shot into a promotion

Place the accepted clip in your editor, add the café name and actual offer as text layers, and finish with a clear action such as a location or ordering link. Choose music separately; the sample has no audio.

For a complete sequence with an offer card and storefront shot, continue with the product video maker. For a reflective object instead of food, see the jewelry photo workflow. The preparation and failure checks differ even though both can start from a single image.