Photo to video: a practical workflow for a subtle moving shot

1. Prepare the source image

Use a sharp JPG, PNG, or WebP image up to 10 MB. Avoid severe motion blur, an obstructed subject, or a crop that already cuts off important details. Match the source composition to your intended video placement where possible; a wide scene and a narrow portrait ask for different framing.

Only upload material you have permission to use. Reference uploads are stored as permanent public file links while the files remain stored, and anyone with a link can access them. Generation sends the media to processing services. Do not use confidential client work as a test image.

2. Write motion, not a new picture

For a lake scene, ask for gentle water ripples and a slow camera push-in. You do not need to describe every mountain already visible in the source. If the prompt introduces a different place, lighting setup, and subject, it competes with the starting image.

Separate stable elements from moving ones. Keep the shoreline and mountain shapes stable while the water moves. This is a direction rather than a preservation guarantee, but it makes the result easier to evaluate.

Gentle ripples move across the lake while a few clouds drift slowly. The camera makes a very slow push-in. Keep the mountain silhouettes and shoreline stable, one continuous shot, no cuts, no new objects, no text.

3. Choose only the inputs the model accepts

Select an image-to-video model and review its upload controls. Some support an optional end frame; Vidu Q4 Preview in this tool accepts one source image without an end-frame option. Do not assume every model accepts the same number of reference files.

Choose duration and resolution, then inspect the current credit quote before submitting. The 5-credit account trial is below current video generation costs. Start with a short test so you can decide whether the motion is useful before paying for another variation.

4. Inspect continuity before downloading

For landscapes, look at the horizon, building edges, water boundaries, and any foreground branches. For portraits, check eyes, teeth, fingers, and changes in facial identity. For products, inspect labels, corners, and reflections. View the beginning and ending frames as well as the motion between them.

If the subject warps, reduce the action and camera movement. If the crop loses an important detail, revisit the source image. Correcting a still image first can be more effective than repeatedly asking the video model to repair the same defect.

5. Prepare the clip for its actual use

Download the completed result from the result area or your library. Confirm dimensions, duration, and any available audio in the actual file. Add approved captions, typography, or music in your editing workflow when required; generated text and audio should not be assumed accurate or available.

Before publishing, check the selected model’s terms and rights to the source and output. Do not imply that an AI-generated event happened at a real location. If you need an entirely invented composition, return to text to video instead of forcing the photo workflow to become a different scene.