Five Locket Photos to Capture Before Using AI
A five-shot locket photography plan covering the closed front, open interior, hinge, reverse and scale before AI creates backgrounds or model images.
A locket is not fully described by its decorated front. Buyers also need to understand what happens when it opens: the interior shape, photo area, hinge, catch, depth and reverse finish. An AI image created from one closed view has no evidence for those details, so any open view it produces is an attractive guess.
The practical solution is a five-photo source set. Photograph the exact locket closed, open, from the side, from the back and at a useful scale before generating backgrounds or lifestyle scenes. This gives shoppers real product evidence and gives an image editor less geometry to invent.
1. Closed front for product identity
Start with the view most shoppers will recognise: the locket closed, facing the camera, with the bail and a section of chain visible. Use soft, controlled light and keep the product large enough to show engraving, stones, borders and the seam between the two halves.
Clean the jewellery first and stabilise the camera. Shopify's current jewellery photography guide recommends a tripod, diffused light and high-quality source files because small details and polished surfaces expose movement, glare and fingerprints quickly.
This closed frame is the identity image. Keep an untouched master and use it to check the silhouette, hinge position, catch position, chain connection and decoration in every later creative. If the front is asymmetrical or engraved, do not flip the image merely to improve a composition. A mirrored locket is no longer an accurate view of the item.
2. Open interior for the usable space
Photograph the real locket fully open and as square to the camera as the mechanism allows. Both halves, retaining rims and any photo windows should be visible. If one side holds a photograph and the other carries an engraving or decorative panel, show that layout clearly.
Do not insert a generated portrait into the only interior image. First publish a clean view of the empty or supplied interior so the customer can see what is included. A styled example can follow as a separate image if it is clearly illustrative.
Record the usable photo opening, not just the outside width of the locket. A simple dimension graphic can be added after photography, but the measurement must come from the physical item. The rim, hinge and retaining frame can make the interior opening noticeably smaller than the outer shell.
3. Side view for the hinge, catch and thickness
Turn the open locket to a three-quarter side angle. Focus on the hinge barrels, pin, closing edge and overall thickness. Then capture the locket closed from the same side if the catch or seam is hidden in the open position.
This is where a generated view is most likely to fail. An image model may create a plausible hinge with the wrong number of barrels, put the catch on the opposite side, remove a gap needed for movement or make the two halves different thicknesses.
Use the real side image for any claim about construction or operation. AI can help clean the background around it, but it should not rebuild the hinge itself. A layer mask is useful because it keeps the photographed product intact while the surrounding pixels change. Adobe's guide to Photoshop layer masks explains how masks hide or reveal areas without deleting the underlying image.
4. Back view for condition and customisation
Photograph the reverse even when it looks plain. The back may show a hallmark, maker's mark, engraving area, wear, plating condition, decorative pattern or the external part of the hinge.
For personalised lockets, separate the blank-product image from the customer's proof. The listing should show where engraving goes and the approximate safe area, while the individual proof should show the exact submitted text. Do not use a generic AI rendering as evidence that a particular name, date or photograph has been checked.
Vintage and pre-owned pieces need even stricter handling. Scratches, dents, repairs and worn plating are part of the condition of the exact item. Remove temporary dust, not information a buyer needs to assess.
5. Scale and wearing position
Finish with a photograph that answers how large the closed locket looks when worn. Use a real person, a neck form or an in-hand scale view, and record the physical dimensions in the listing. The chain length, pendant drop and bail all affect where the locket sits.
If you generate a model image, provide the locket dimensions and chain length, then compare the result with the real scale frame. Our guide to accurate necklace try-on explains why pendant position cannot be judged from appearance alone.
Treat the chain and clasp as part of the SKU. The related guide on preserving jewellery chains and clasps covers the repeated links and small hardware that image models often simplify.
Use AI around the locket, not instead of these views
Once the five-source set is approved, AI can be useful for lifestyle backgrounds, seasonal crops and advertising concepts. Begin each independent version from the relevant real photograph. A closed lifestyle image should start from the closed master; an open gift image should start from the real open interior.
Do not ask an image-to-video model to animate the locket opening unless you also have real motion footage for comparison. A smooth video can still morph the hinge, change the interior or move the catch between frames. For a mechanism, continuity is not the same as accuracy.
When replacing a background, mask the complete product, including the chain gaps and the space around an open hinge. The workflow for putting a real product into an AI background is safer than prompting a model to recreate the entire locket in a new scene.
The final locket check
Compare every edited image with the relevant source at normal listing size and at high magnification.
The hinge remains on the correct side and has the same construction.
The catch, seam, bail and chain connection have not moved.
The two halves retain their true thickness and opening angle.
The interior shape and retaining rims match the physical item.
Engraving, photographs, hallmarks and decoration are not mirrored.
Chain length and worn scale remain believable.
Condition marks have not been hidden on one view and shown on another.
The goal is not to avoid AI. It is to stop AI from becoming the only source for a part of the product that was never photographed. Five accurate views give the shopper a complete locket and give creative tools a boundary they can work around.