Prompting for Images & Video

Prompting for Images & Video

Writing prompts for image and video generators is a distinct skill from prompting a chatbot. Conversational AI rewards nuance and natural explanation. Visual AI rewards precision, specificity, and the right vocabulary.

The Anatomy of an Image Prompt

Subject — Be specific. "A woman" produces very different results than "a middle-aged woman with short silver hair wearing a navy coat, standing on a cobblestone street."

Style — Photorealistic, oil painting, watercolor, pencil sketch, anime, cinematic, illustration.

Lighting — "Golden hour," "dramatic side lighting," "soft diffused light," "neon glow," "overcast" all produce distinctly different atmospheres.

Camera or perspective — "Wide angle," "telephoto," "macro," "aerial view," "eye level," "portrait orientation."

Quality modifiers — "Sharp focus," "highly detailed," "professional photography," "8k."

A complete prompt: "A middle-aged woman with short silver hair, standing on a cobblestone Paris street, cinematic lighting, golden hour, wide angle lens, photorealistic, highly detailed."

Negative Prompts

List what you do not want. Common entries: "blurry, distorted, extra limbs, watermark, text, low quality." Particularly useful for avoiding AI's well-known struggle with hands — add "deformed hands, extra fingers" to the negative prompt.

How Different Tools Respond

• Midjourney — Responds well to both natural language and keyword-heavy prompts. Strong aesthetic bias that can override specific instructions.
• DALL-E 3 — Works best with descriptive sentences. Trained to follow complex written instructions closely.
• Flux — Favors descriptive, keyword-rich prompts. Excels when specific about subject matter and style.

Seed Numbers for Iteration

When you find a result you like, note the seed number. Keeping the seed constant while changing one prompt element lets you make controlled comparisons.

Prompting for Video

Video prompts add temporal and motion elements: - Camera motion: "slow pan left," "dolly forward," "handheld shake," "crane shot descending" - Subject motion: describe what moves and how — "leaves falling slowly," "a figure walking toward the camera"

Iteration Is the Process

No prompt produces the perfect result on the first try. Effective use of visual AI is iterative: generate → evaluate → adjust one or two variables → generate again. Keeping notes on what worked builds a personal library of techniques.