Prompting for good images and video

Getting good AI images and video comes down to describing what you want clearly — and there's a simple recipe anyone can follow. Here's how to prompt for good results.

The image prompt recipe — describe these:

• Subject — who or what: "an older fisherman mending a net."
• Style or medium — "watercolor," "35mm film photo," "3D render," "cinematic," "oil painting."
• Composition / shot — "close-up," "wide landscape shot," "top-down flat lay," "portrait."
• Lighting / mood — "golden hour," "soft studio lighting," "moody and foggy," "bright and cheerful."
• Extra details — color palette, era, aspect ratio, camera/lens feel.

You don't need all of these every time, but naming subject + style + composition + lighting gives the model a clear picture and dramatically improves results over a bare "a fisherman."

Core skills (some familiar from prompting in general):

• Iterate. Treat the first result as a draft — change one thing at a time and regenerate (results vary each run anyway). Iteration is how you home in on what you want.
• Use reference images. Most tools let you upload an image to guide the style, composition, or a character's look. Powerful for consistency — but only use references you have the right to use.
• Be specific but not overloaded. A clear, focused prompt usually beats a very long, contradictory one. Add detail deliberately, not randomly.

For video, add motion:

• Describe the camera movement — "slow push-in," "drone shot flying over," "static shot."
• Describe how the subject moves — "gently turning to look at the camera."
• Keep clips short and simple — AI video is most reliable on brief, focused scenes; complex multi-action sequences are harder.

Manage expectations. Even with great prompts, expect some misses — hands, fine text, exact object counts, and keeping a face consistent across shots are still tricky (last lesson). Generate a few options, pick the best, and iterate. Getting a great result is normal to take several tries — that's the craft, not a failure.

The mindset: good AI images and video come from a simple recipe — describe the subject, style/medium, composition, lighting, and key details (and for video, the camera and subject motion, keeping clips short). Then apply the core skills: iterate (change one thing at a time and regenerate, since every run varies), use reference images you have rights to for consistency, and be specific but not overloaded. Manage expectations — hands, text, counts, and consistent faces are still hard, so generate options and refine. Clear description plus iteration is the craft that turns AI image tools from a novelty into a creative instrument.