Hi everyone! I have a task to build a workflow for mass processing of product images (scale — hundreds of thousands of photos).The goal is to automatically cut out individual infographic elements (icons, the main product, zoom-in areas) and bring them to “premium” quality (studio lighting, high sharpness). Problem with GPT Image 1.5 When trying to directly cut out an object using GPT Image 1.5, the model starts to hallucinate. It redraws small details: Faces on icons change LED diode grids get distorted Fabric textures are altered As a result, the output image is no longer the same product as in the original photo. Either the model changes the appearance of the product, or the visual details become inaccurate. Are there any ways to implement this differently, or alternative approaches/models that can solve this task without such distortions?