← ClaudeAtlas

product-photo-from-referencelisted

Generate new product photography from a real product photo using reference_images, so the product stays accurate instead of being reinvented. Use whenever the user has a product photo and wants more shots, new angles, new scenes, or variations, or asks how to get consistent product imagery.
apimageorg/apimage-skills · ★ 0 · AI & Automation · score 72
Install: claude install-skill apimageorg/apimage-skills
# Product Photos from a Reference The core APImage image workflow: one real product photo in, many accurate variations out. It replaces a photoshoot for scenes, angles and settings — but only if the product stays the product. Generating a product from a text description produces something that *resembles* yours. For commerce that's not a stylistic difference, it's advertising a thing you don't sell. ## Always start from the real photo ``` generate_image( model="flux-2-pro", reference_images=["<real product photo>"], prompt="The same bottle, on a pale oak table, morning window light " "from the left, shallow depth of field, warm neutral grade. " "Product unchanged — same label, same shape, same colour.", aspect_ratio="1:1", seed=8812 ) ``` `reference_images` accepts **up to 4 for images**. Use all of them for anything where accuracy matters. Three clauses doing work: **"The same bottle"** rather than describing the bottle. The reference supplies the product; re-describing it competes with the reference and causes drift. **"Product unchanged — same label, same shape, same colour."** It won't fully prevent drift but it measurably reduces it. **Everything else** is the scene — surface, light, grade, depth. That's what you're actually generating. ## The reference photo decides the ceiling A mediocre reference produces mediocre variations and no prompt fixes it. **What makes a good reference:** - Sharp, evenly lit, product filling most of the