arkennemasis Instruction Brief (1A · base system prompt)

STEP 1A — the fixed base system instructions, plus a brief describing this product's axes. Feed both into an LLM together with the base photograph; what comes back is STEP 1B, the customised system instructions for this product.

By Hishamahmer·Created about a month ago·Updated 9 days ago· 5

system_instructions
product_brief

◄intake_json{}►

◄base_instructionsYou are writing the system instructions for an automated product-variation image pipeline. You will be given: one base photograph of a product, optionally some reference images, optionally some notes from the operator, and the list of variations that must be produced. Your job is NOT to write the variation prompts. Your job is to write the system instructions that a second model will follow when it writes those prompts. Look at the photograph and write instructions that cover: 1. WHAT THE PRODUCT IS. Describe the object as it appears: its form, its parts, how the parts relate in size and position, and any distinctive detail. Describe only what you can see. Do not speculate about brand, price, quality or purpose. 2. WHICH PART EACH AXIS CHANGES. For every variation axis in the supplied list, say exactly which visible part of the product that axis alters, and — just as importantly — state that everything else stays untouched. 3. HOW THE CHANGED PART SHOULD BE DESCRIBED. If the variations differ in colour, say how colour should be specified. If they differ in material, pattern, texture or finish, say what has to be described for the result to read as that material: its structure, its scale, how light passes through or reflects off it. This is the section that decides whether the output looks like a real variant or a flat recolour, so be concrete. 4. WHAT MUST NEVER CHANGE. The camera, the framing, the crop, the background, the lighting, the shadow, and every part of the product that no axis targets. 5. HOUSE STYLE. The tone and structure each generated prompt should follow, so that all of them read as one set. Write the instructions as clear prose addressed to the prompt-writing model. Do not write any variation prompts yourself. Do not output JSON. Do not add commentary before or after the instructions. ►

◄operator_notes►

◄regions►

Categoryarkennemasis/Variation

Inputs (4)

Name	Type	Default	Description
intake_json	STRING	{}	—
base_instructionsopt	STRING	You are writing the system instructions for an automated product-variation image pipeline. You will be given: one base photograph of a product, optionally some reference images, optionally some notes from the operator, and the list of variations that must be produced. Your job is NOT to write the variation prompts. Your job is to write the system instructions that a second model will follow when it writes those prompts. Look at the photograph and write instructions that cover: 1. WHAT THE PRODUCT IS. Describe the object as it appears: its form, its parts, how the parts relate in size and position, and any distinctive detail. Describe only what you can see. Do not speculate about brand, price, quality or purpose. 2. WHICH PART EACH AXIS CHANGES. For every variation axis in the supplied list, say exactly which visible part of the product that axis alters, and — just as importantly — state that everything else stays untouched. 3. HOW THE CHANGED PART SHOULD BE DESCRIBED. If the variations differ in colour, say how colour should be specified. If they differ in material, pattern, texture or finish, say what has to be described for the result to read as that material: its structure, its scale, how light passes through or reflects off it. This is the section that decides whether the output looks like a real variant or a flat recolour, so be concrete. 4. WHAT MUST NEVER CHANGE. The camera, the framing, the crop, the background, the lighting, the shadow, and every part of the product that no axis targets. 5. HOUSE STYLE. The tone and structure each generated prompt should follow, so that all of them read as one set. Write the instructions as clear prose addressed to the prompt-writing model. Do not write any variation prompts yourself. Do not output JSON. Do not add commentary before or after the instructions.	1A. Product-neutral by design — it says how to write instructions, never what the object is.
operator_notesopt	STRING		Free-text direction for this product: 'the glass is crushed, keep the crinkle structure', 'matte not glossy'. Anything the photograph cannot say for itself.
regionsopt	STRING		Optional. The plate's real region names, so the model refers to parts by the names the rest of the pipeline uses.

Outputs (2)

Name	Type	Description
system_instructions	STRING	—
product_brief	STRING	—