← ClaudeAtlas

product-in-hand-shotslisted

Generate products held in hands, which communicates scale and use — and handle the fact that hands are what generative models get wrong most. Use whenever the user wants a product held, an in-use shot, hands in frame, or a scale reference via a hand.
apimageorg/apimage-skills · ★ 0 · AI & Automation · score 72
Install: claude install-skill apimageorg/apimage-skills
# Products in Hand A hand holding a product does three jobs at once: it establishes scale, it demonstrates use, and it adds human presence. That's why it's one of the highest-converting gallery images. It's also the shot most likely to produce a six-fingered hand. ## Why hands are hard Hands have many joints, they self-occlude, they deform, and the model has seen them in every configuration. The result is the well-known failure mode: extra fingers, missing fingers, joints bending the wrong way, thumbs on the wrong side, fingers merging into the product. None of it is fixable by prompting harder. It's mitigated by **giving the model less hand to get wrong.** ## The mitigations, in order of effectiveness **1. Crop the hand.** A hand entering frame from the edge, cropped at the wrist or mid-forearm, has far less visible structure than a full hand on a plain background. **2. Wrap the fingers.** Fingers curled around the product hide most of the joints. Splayed fingers expose every one of them. **3. Occlude with the product.** A hand mostly behind the product shows only what's needed. **4. Keep it out of focus.** A hand at the edge of a shallow depth of field is softer and more forgiving. **5. One hand, not two.** Two hands doubles the failure surface and adds the problem of them relating correctly. **6. Just photograph it.** For a critical gallery image, one real photo of a hand holding the product removes the entire risk class. Then generate scenes around it with `re