craft-image-promptslisted

# Craft Image Prompts Turn an approved visual brief into a concise, executable prompt. Speak French unless asked otherwise; write the final image prompt in the language and format requested (English by default). ## Establish the brief Infer what is already clear. Ask no question when a safe, conventional assumption is sufficient. Otherwise ask at most three high-value questions, prioritised as follows: 1. Task: generate, edit, composite, variation, or data visualisation. 2. Required output: target model, aspect ratio/dimensions, and any text that must appear in the image. 3. Creative intent: subject/action, medium or visual style, reference-image role, and non-negotiable constraints. If the target model is unknown, make a model-neutral prompt; do not falsely claim that its syntax is optimised for a named model. Use web research only if the user asks for a current model-specific syntax or capability check. Internally extract: subject hierarchy, scene/environment, composition, palette, lighting, material/texture cues, physical constraints, text content, preservation requirements, forbidden elements, references, and output settings. ## Choose a prompt strategy Classify the request before writing: | Request | Include | | --- | --- | | Generation | Asset type, subject/action, environment, composition, relevant rendering details, constraints. | | Editing | Exact change, elements to preserve, continuity of lighting/perspective/materials, local scope. | | Reference-led | Wha