· Updated 11 November 2026 · 13:30 CET
How to turn a still image into a short video
Animate one product image with two camera directions, compare the clips and choose the movement that keeps the product clear.

The bottle shop has a product image and needs a short moving shot for its page. We will use the same still for two attempts, asking for a different camera movement in each. The task is to keep the bottle recognisable and fully visible.
1. Pick a simple starting image
Choose a sharp image with a clear subject and enough room around it. This is the actual starting image supplied to both video generations:

2. Try a gentle push-in
Open Create a video and choose an image-to-video model. Upload the still as the starting image. For these examples we used Kling v3 Video via Replicate, standard mode, three seconds and audio disabled. No last frame or additional references were supplied.
A slow gentle camera push toward the sage-green bottle. The bottle stays upright and still, with its shape and steel cap unchanged. Soft studio light, calm product film. No text, no new objects.
Watch attempt 1 — camera push-in

3. Change the movement
The push-in takes the camera too close for a complete product view. We used the same still and settings for a second attempt, changing the camera instruction:
A very small smooth camera slide from left to right past the sage-green bottle. The bottle stays upright and still, with its shape and steel cap unchanged. Soft studio light, calm product film. No text, no new objects.
Watch attempt 2 — gentle sideways slide

4. Choose and finish the shot
For this product-page task, the sideways version is the more useful final choice because the whole bottle remains visible. Play both clips and watch the cap, edges and shadow for changes. A prompt that asks to preserve shape still needs a visual check.
Download the chosen clip, trim it to your edit and add captions or audio in your video editor if needed. These are two actual generated clips, not still images presented as video. Their frame size follows the supplied image, and both were requested at three seconds.
Model inputs: Kling v3 Video on Replicate.