Skip to main content
  • Pricing
AI image

Nano Banana Dual-Reference Test for Jewelry Scale

By Oakgen TeamPublished 1 min read

Share
Nano Banana Dual-Reference Test for Jewelry Scale

A product reference provides details while a different reference may specify fit and scale. If the two signals conflict, use physical labels and measuring points in the review. For jewelry visualization, a believable styling pass can be misleading if it visibly changes the size of the stone or setting.

Build the test input

Use a round original ring with a stone and a tabletop reference marked with ruler lengths. Prepare a separate authorized hand photo showing scale, with the fingers relaxed. Decide which image controls geometry and which controls placement. Avoid adding pose or light cues that contradict the fit reference.

Run the proposed comparison

Compare a single combined input with a labeled pair where supported. Preserve the exact input order and same output crop for each trial. Trace the stone diameter and finger width from the approved hand photo, then calculate their ratio in each output. Avoid measurements on frames with the stone turned edge-on.

Decide what passes

Flag an apparent ratio significantly above the ratio you approved for the brief. Confirm that the stone count and setting orientation are preserved too. An incorrect fit must fail even if lighting and skin look natural, because the output can give viewers a false expectation about the piece.

Act on the outcome

For design review, a model pass may reveal the intended fit visually. For commerce photographs, prefer approved packshots or manual composite when the model invents small details. Check another ring size separately before scaling a workflow beyond the specific source and hand used here.

Model references

Share