Help AI understand the image in your head.
You have a campaign idea. Before ChatGPT or Claude can turn it into a useful generation prompt, it needs to understand what you want the finished asset to look like.
If you leave out the details, AI has to fill in the gaps. It might choose the wrong colors, change the product’s finish, or arrange the scene differently from what you imagined. That misunderstanding carries into the prompt you use on Higgsfield or another image or video platform.
Then you spend credits generating something you did not want, revise the prompt, and pay to generate it again. Clear art direction helps reduce that avoidable trial and error. It does not guarantee a perfect result, but it gives you a more specific starting point.
You do not need to know photography or design terminology. Describe what you can in everyday language. Ask ChatGPT or Claude to help you define the rest, show you options, and explain what each choice will look like before it writes your prompt.
1. Explain the idea and what the asset needs to do.
Start with the product, the campaign idea, the feeling you want, and where the asset will appear. Explain what you want someone to notice first.
Instead of only saying “make it premium,” describe what that means to you: a quiet setting, a glossy surface, a close view of the product, or a playful interaction with a prop. You can use reference images to help explain your taste.
2. Define the product, colors, and finishes.
Attach a clear product photo. Tell AI which details must stay accurate: shape, proportions, packaging color, cap or dispenser, label, and material. Identify which reference shows the product and which only provides inspiration.
Describe the colors and surfaces of the props and background separately. A surface might be matte, shiny, brushed, mirrored, or textured. If you want a prop to borrow a packaging color, say which prop and which packaging reference.
If you do not know the name of a finish, point to it in an image and ask AI to describe it. Ask it to flag details it cannot verify.
3. Decide where everything goes.
Composition simply means how things are arranged inside the frame. Decide whether the product stands upright, lies horizontally, or sits at an angle. Describe its position, size in the frame, and relationship to any props.
Think about the view: straight on, from above, or from a slight angle. Decide whether you want a close crop or the entire scene, and whether you need empty space for text.
You can say, “I want the bottle standing up in the middle, with the full prop visible beside it.” ChatGPT or Claude can translate that into more precise composition instructions.
4. Describe the light and the overall feel.
Lighting changes how a product’s shape, texture, and finish appear. Describe whether you want soft shadows or sharp ones, bright reflections or a subdued surface, and a warm, cool, or neutral overall color.
If you like the lighting in a reference, tell AI to analyze it and turn it into instructions. You can also ask, “What lighting would help me get this feeling?” Have it explain the options without assuming you know studio terminology.
5. Approve the direction, then ask for the prompt.
Ask ChatGPT or Claude to bring the decisions together into one short brief. Read it as though you are describing the finished picture to someone who has never seen your idea.
For video, add the duration, what moves, how it moves, what stays still, whether the camera moves, and any sound. Keep the action simple enough to describe clearly.
Once the brief matches your idea, ask for a prompt for the platform and model you plan to use. Review the result against your references before spending more credits. If something is wrong, name the specific correction and the approved details to preserve.
“I have an idea, but I do not know how to describe the art direction. Help me define the product placement, colors, finishes, composition, and lighting. Ask me questions, explain the options simply, and write the generation prompt only after we agree on the direction.”