How to turn a product photo into a video ad with AI
Start frames, end frames and motion prompts: how to animate a product photo with Kling 3.0, Seedance 2.5, MiniMax H3 Max or Gemini Omni Flash 1.1.
Short answer: open Video Generate, drop the product photo into the Start frame, and describe what moves and how the camera moves. An End frame is optional. Pick a model, check the cost on the Generate button, and the clip lands on the board, starting from your product as it looks in the photo.
Starting from a still is the simplest way to keep a product true in motion. A prompt alone has to invent the bottle, the label and the colour; a start frame hands the model the real thing, so the clip only has to decide what happens next.
What makes a good start frame?
- The product, clearly. The model continues what it is shown. A clean, well-lit photo of the product carries through better than a busy one.
- The shape of the placement. Pick a photo that already suits the ratio you will run, vertical for Stories and Reels, wide for a banner, so the frame does not have to be cropped around the product.
- Room to move. If the camera is going to push in or orbit, leave space around the product for it to travel.
No photo yet? Generate one first in Image Generate. A still on the board opens in Video Generate with Animate.
How to turn a product photo into a video, step by step
- Open Video Generate and drop the photo into the Start frame.
- Write what moves: the product turning, liquid pouring, a hand reaching in. Then say how the camera moves and how the light falls.
- Add an End frame if the clip has to land on a particular shot, such as the packshot with the logo facing out.
- Pick the model, then set the length, the ratio and, on models that make it, the sound.
- Press Generate. The cost is on the button before you press it, and the clip lands on the board.
Which model should animate the photo?
| Model | From a photo | Clip length | Sound |
|---|---|---|---|
| Kling 3.0 | Start and end frames | 3–15 s | Yes |
| Seedance 2.5 | One or two images as start and end frames; three to thirty as references | 4–30 s | Yes |
| MiniMax H3 Max | Start and end frames, or up to 9 references | 5–15 s | No |
| Gemini Omni Flash 1.1 | Start and end frames, or up to 10 references | 3–10 s, up to 4K | No |
Seedance 2.5 suits a longer product story, Kling 3.0 a clip with sound that lands on a set end frame, Gemini Omni Flash 1.1 a 4K hero shot, and MiniMax H3 Max quick openers made in volume. The video model comparison goes through each one in more detail.
How do you write the motion prompt?
Describe the change, not the photo. The model can already see the product; what it needs is the action, the camera and the light. “The bottle turns slowly to face the camera, the camera pushes in, warm light sweeps across the glass” gives it all three. Keep one main action per clip: a short ad reads better as one clear movement than as several at once.
What does the clip still need before it ships?
A silent clip needs a voice or a track: Text to Speech reads the script, and Music Generate writes a bed. Feeds often start muted, so Caption adds subtitles. Video Reframe recuts the clip for other placements, and Video Extend carries it on past its last frame when it ends too soon.
Questions people ask
Can I use my own product photo?
Yes. Drop it into the Start frame and describe what moves.
Do I need an end frame?
No. It is optional. Add one when the clip has to finish on a particular shot.
Which models add sound to the clip?
Kling 3.0 and Seedance 2.5 generate sound in the same run. MiniMax H3 Max and Gemini Omni Flash 1.1 return silent clips.
Can I animate an image I generated in MarketingX?
Yes. A still on the board opens in Video Generate with Animate.
Put your first product photo in motion in Video Generate.
