The step that puts your product onto the photo somebody sends: choosing the product image, the nine product kinds, the ready-made prompt, the waiting and retry settings, the three outputs and the limits.
Last updated: September 30, 2026
"Would these glasses suit me?" is a question no photo album answers. The Try it on step answers it directly: the person sends a photo of themselves, AI puts the product you chose onto it, and the result goes straight into the conversation. This guide covers how the product image is chosen, what the nine product kinds are for, every waiting setting, and the three outputs.
In this article you will find:
⚠️ Note: The step spends AI credits and needs a key that can generate images. If your plan does not cover AI, there is no image-capable key, or credits have run out, the step does not run and the flow continues from Could not generate.
The step combines two images: the person's photo and the product image you chose. The result is a new photo with the product on them.
One guarantee is given up front and is written into the ready-made prompt: the person's face, expression, skin tone, hair, body proportions, posture and background do not change. Only the product is placed. Left without instructions, image models "beautify" people — slimming them, lightening skin, fixing faces — which would defeat the whole point of a try-on.
The person's photo is never stored: it is downloaded at call time, given to the model and dropped from memory.
The product to be tried on. Use Choose from library to pick one of the files in your media library and Change to swap it for another.
You cannot type an external address — that is a deliberate restriction. The product's bytes are read from the database and no request goes out.
If no product image is chosen the check raises a warning (the flow can still be published; at run time the step leaves through Could not generate). An address that is not from the library is an error and blocks publishing.
💡 For the best result, supply the product on a plain background, on its own, well lit. A photo of the product on a model confuses the system.
The choice that decides the ready-made prompt. There are nine kinds: Clothing, Jewellery, Bracelet, Watch, Glasses, Shoes, Bag, Hat and Other.
This fixes what models get wrong most often: where on the person the product goes and at what scale. A bracelet belongs on the wrist, glasses on the nose and ears, a hat on the head. Other works for any product but leaves the placement to the model.
The text that asks the person for their own photo. It is mandatory: if empty, the pre-publish check reports an error — the step does not move on until something arrives, and the person has no idea what to send.
Example: "Send a photo of yourself and I'll show you the product on you ✨"
On by default. If the person already sent a photo in the message that started the flow, the step skips the question and goes straight to generating. Turn it off to always ask for a photo.
A short message sent before generation starts. Generation takes seconds; this keeps the person from sitting in silence. May be left empty.
Example: "Great, working on it… a few seconds 🙌"
⚠️ Note: If this message cannot be sent, generation is not attempted at all and the flow continues from Could not generate. The reasoning is simple: if a message will not go out, the result image will not either, so no credit should be spent.
The text sent with the generated image. May be left empty.
The instruction handed to the model. It comes ready from the product kind and the field stays empty. Use Edit to write your own and Reset to default to go back.
Leaving the field empty is deliberate: when the ready-made prompt is improved, every flow that has not touched it picks up the new text automatically.
Writing your own prompt is a separate plan feature; working with the ready-made prompt is the step itself.
How long the person's photo is waited for: between 1 and 168, default 24. When the time runs out the flow continues from No answer.
⚠️ Note: The check warns for windows over 24 hours. Free-form messages cannot be sent more than 24 hours after the person's last message, so the result image cannot go out in that window.
If the person sends text or another kind of attachment instead of a photo, how many more times the photo is requested: between 0 and 3, default 1. When the allowance runs out the flow continues from Could not generate.
The text sent when the photo is requested again. Left empty, the original question is repeated.
Connect all three. With Could not generate and No answer empty, somebody who wanted to try a product gets nothing back; the check lists both as warnings.
The step has two turns:
If the provider is busy the run is deferred; this happens at most twice, after which it moves to Could not generate. The deferral limit is deliberately low: the person is waiting in the conversation, so the prepared message you wrote should step in rather than the run dragging on.
An optician is promoting a new model on Instagram.
try.The result: the person who commented sees the glasses on their own face within seconds.
Trying to supply the product image by typing a website address. It can only be chosen from the library; a typed external address is an error and blocks publishing.
Leaving the product kind as "Other". Placement is left to the model, and a bracelet may end up around a neck. Pick the right kind.
Leaving the photo question empty. Publishing is blocked, and even if it were not, the person would not know what to send.
Making it the first step of a comment flow. The first message out of a comment carries plain text only, so the result image is dropped. Put a message step in between.
Setting the wait to 168 hours. The result image cannot go out once the 24-hour window has closed.
Leaving the Could not generate branch unconnected. The cases where the model says "I can't" are real: with no person in the photo, or the relevant area hidden, no image is produced.
Expecting good results from crowded, dark or very distant photos. Tell the person what kind of photo to send — "with your face clearly visible", "full length".
For text-only jobs, see The AI step. To send a ready product image as it is, The Media step is enough. What happens when credits run out is covered in AI credits have run out.
Once somebody sees the product on themselves, "would it suit me?" stops being a question — and that is usually when the buying decision is made.
Was this article helpful?