How do you generate AI video from a photo?

Upload the photo to an image-to-video tool, write a short prompt describing only the motion you want (not the scene — the scene is your photo), and generate. The model uses your photo as the first frame and animates from there. Best image-to-video models in 2026: Veo 3, Kling 2 and Runway Gen-3.

Prompt motion only, not the scene.
Veo 3 / Kling 2 / Runway Gen-3 lead image-to-video in 2026.
Higher-resolution input photos give better results.
Square-on photos work better than extreme angles.

The prompting trick most beginners miss

With text-to-video you describe the whole scene. With image-to-video, the scene is already locked by your photo — the model just needs to know what should move. Prompt only motion: 'slow camera push-in, subject smiles, wind blows hair gently'. Re-describing the scene confuses the model.

Photo quality matters

The output cannot be higher quality than the input. Use a clean, well-lit photo of at least 1080×1920 for vertical or 1920×1080 for horizontal. Avoid heavy filters or compressed Instagram screenshots — the model amplifies the artifacts.

FAQ

Can I use any photo I want?

Photos of yourself, your products, or licensed stock = yes. Photos of other real people or copyrighted characters = legally risky and against most platform terms.