Practical image-to-video guide
Prepare a strong source, write a focused motion prompt, choose a model, and iterate safely.
A clean source gives the motion model fewer visual conflicts to solve. Start with synthetic artwork or material you have permission to use.
Decide what the viewer should notice first. One main action is easier to preserve than several competing movements.
Describe change over time. The image already carries appearance, clothing, lighting, and composition, so use the prompt to direct movement.
Use a fast model to test the direction. Move to a higher-control or reference model when the source and prompt are already working.
Evaluate subject fidelity, anatomy, motion, camera behavior, and unwanted scene changes. Adjust one variable per retry.
Use these structures as starting points, then replace the bracketed details with your scene.
Best when preserving appearance matters more than a large action.
[subject] makes a slow, natural [movement]. Subtle breathing and hair motion. Camera remains steady. Preserve face, anatomy, lighting, and background.
Use one camera action that supports the subject instead of competing with it.
The subject holds the pose with gentle natural motion. Slow camera push-in. Keep proportions, expression, outfit, and scene layout consistent.
Add life to a mostly stable pose through lighting, fabric, or background detail.
Minimal subject movement. Soft fabric and hair motion, subtle light shift, calm background movement. Locked framing and consistent character details.
A short review protects source rights and reduces avoidable failed tasks.
Every depicted person must clearly be an adult. Content involving minors is prohibited.
Use synthetic, illustrated, self-owned, or explicitly permitted source material.
Do not create sexualized content of celebrities or other real people without valid consent.
Inspect the output before saving, exporting, or using it in another workflow.
Quick answers for common image-to-video problems.
Small movements give the model less new anatomy and scene geometry to invent, which often improves source fidelity.
Use them only when the selected model supports them. Prioritize a clear positive motion direction first.
Change one main variable when possible. This makes it easier to identify whether the source, motion prompt, camera direction, or model caused the issue.
Open the image-to-video generator, upload a valid source, and test one subtle motion prompt with a fast model.