Collar shape and front placket. Cuffs, sleeve length and buttons. Stripe or check continuity.
Put the collar and placket first
A shirt is familiar enough that a plausible result can escape close inspection. Make the review specific before generating: collar shape, front placket, pocket position, cuff construction and hem. Those details separate your product from a generic shirt in the same color. Choose a source with the front visible rather than a heavily folded retail arrangement.
Decide whether the shirt should be buttoned, open or hanging as it appears in the reference. Changing the styling can hide the button line and make construction harder to compare. For the first sample, a simple product-led view gives you a clearer basis for approval.
Follow the pattern across the seams
Stripes and checks deserve their own review pass. Follow a repeated line across the body, sleeve and pocket rather than inspecting only the center of the image. Look for changes in spacing, direction and continuity. Small irregularities can be easy to miss when the lighting and background feel persuasive.
Use soft light that keeps the surface readable. A hard shadow across the placket may add atmosphere, but it can also prevent you from checking the button spacing. If the material is sheer or textured, retain an original photograph that establishes its actual appearance instead of expecting the campaign image to answer every material question.
Let the placement guide the styling
A hanging shot can communicate the complete garment; an on-model image can add a styling context. Neither proves fit. When choosing a model pose, keep arms from covering the cuff or pocket you need to show, and check that a tucked hem has not turned into a different shirt length.
Compare the sample with your source at full size and as a thumbnail. Once it passes, reuse the direction across related shirts, preserving separate SKU review. If a patterned shirt fails while a plain shirt succeeds, simplify that item's pose or input rather than restarting every accepted result.
