Start with the right request.
Start with an existing product photo when the result needs to represent a real item. Attach it to the request and describe the background, light, and framing you want. A prompt alone is useful for an invented product concept, but it cannot identify a merchant’s actual inventory.
A prompt to build from
Place the attached blue ceramic cup on a pale oak table beside a window. Keep the cup’s shape, handle, glaze and proportions unchanged. Use soft daylight, a natural contact shadow and a quiet cream background. No extra objects or text.
Example prompt, not a promise of exact results. Review the generated image before publishing.
Send it to the Image API.
POST this JSON to /api/v1/images/generations with your Magnt key in the Authorization header. Replace the image placeholder with your actual JPEG bytes encoded as base64.
{
"model": "magnt-image-1",
"prompt": "Place the attached blue ceramic cup on a pale oak table beside a window. Keep the cup’s shape, handle, glaze and proportions unchanged. Use soft daylight, a natural contact shadow and a quiet cream background. No extra objects or text.",
"images": [
"data:image/jpeg;base64,YOUR_IMAGE_BASE64"
],
"aspect_ratio": "1:1",
"resolution": "1K",
"response_format": "image"
}Review the details that matter.
Compare the result with the original at full size. Check logos, labels, proportions, colour, material and the number of items shown. Generative editing can change details even when the prompt asks it not to. Treat the output as a candidate for review, not a verified record of the product.
Make it part of your app.
Let the merchant choose a scene, generate one preview, and explicitly accept the result. Magnt returns the image bytes; your app decides whether to download or save the accepted image. There is no image library to synchronise with Magnt and no stored product context carried into the next request.
Can I generate a whole catalogue in one request?
Each request returns one image. There is no batch endpoint, and the current account limit is 30 generation requests per hour. Start with a user-triggered preview workflow rather than a large automated catalogue import.