You find the outfit, you screenshot it, you upload it, and Snagfit comes back with nothing. That is annoying, and almost every time it is the photo rather than the clothes.
Snagfit reads each garment in a picture on its own: the jacket, the trousers, the shoes, the bag, each identified separately along with its color, cut and material. Every garment it finds is then turned into a short shopping search, four to eight words long, built from the type of garment, its color and any print on it. So the picture itself is never searched. A description of the picture is. That is the useful part, and it is also what sets the rules below. A garment the AI cannot see clearly is a garment it cannot describe, and a garment it cannot describe is one it cannot match.
How much of the outfit should be in the frame?
A full-body shot gives the AI the most to work with. When a photo is cropped at the waist, the trousers and shoes are simply not there to find, so the results come back short and it looks like a miss.
This matters most with video. Dragging the video to roughly the right moment usually lands you on a frame where an arm, a turn, or a passing object is covering half the outfit. Move a second either way and take the screenshot there instead.
Shoes are the piece people lose most often. They sit at the bottom of the frame, which is exactly where a screenshot gets cut by a play bar or a caption. If the shoes are what you came for, check they are fully in the picture before you upload it.
Why does a blurred frame fail?
A frame grabbed mid-movement is blurred, and blur removes exactly the detail the AI is reading: the weave, the seam lines, the edge where one garment ends and the next begins. A still moment works better than a dramatic one.
What if something is covering the clothes?
Bags, arms, furniture, other people. Anything in front of a garment reads as part of the shape. If the coat is half behind a counter, expect the coat to be the piece that comes back wrong.
What happens with several people in the photo?
When several people are in the picture, the AI works on the main subject: the person the photo is about. It does not return a separate list for everyone in frame.
That matters if the outfit you want belongs to someone standing off to the side. Crop the picture down to that person first, then upload the crop. Your phone's photo editor is enough for this, and cropping costs you nothing, because the picture gets resized before it is sent anyway.
Do filters change the result?
Color is part of every match. The AI writes down the main color of each garment in plain words, and that color goes straight into the shopping search. A strong filter, or a room lit by one strongly colored light, can turn a green coat grey and send the search looking for the wrong coat.
Prints suffer in the same way. A soft filter can smooth a check pattern into a flat block of color, and the pattern is often the thing that makes a piece findable. Given a choice of frames, take the one where the clothes look closest to how they would look in daylight.
What does not matter at all?
A few things people worry about that make no difference:
- File type. JPG, PNG and WEBP all work, and a plain phone screenshot is already one of those.
- File size. Large photos are resized inside your browser before anything is sent. The longest side is brought down to 1280 pixels and the image is saved as a JPG, so a big camera roll photo uploads at about the speed of a small one.
- Where it came from. A screenshot from a video, a post, a friend's message or your own camera all work the same way. Snagfit reads the image, not its history.
What happens to the photo you upload?
The photo is sent to the detection service to be read. It is not published, and it is not shown to other visitors. Your recent fits are saved against your own session so you can open them again later, and each one has a delete button on it. The FAQ goes into more detail, and the privacy section is the full version.
How do you frame for one piece rather than the whole look?
The advice changes depending on whether you want everything or one thing, and most people never make that decision consciously.
If you want the whole outfit, keep the full figure in frame, head to shoes. Snagfit returns up to eight garments and can only describe what is visible, so anything cropped out simply does not appear. The cost is that each individual piece is smaller, and the descriptions are less specific.
If you want one piece, crop until that piece fills most of the picture. The description you get back becomes far more detailed, because the collar, the closure, the hem and the fabric texture are all now readable.
If you want both, take two screenshots. One wide for the inventory, one tight for the piece you care about. It takes twenty seconds and removes the tradeoff entirely. This is the habit that separates good results from frustrating ones, and it is covered properly in the post on cropping.
Why do small items need their own photo?
Some garments are almost never findable from a full-body frame, and it is worth knowing which before you blame the search.
Shoes are the clearest case. In a normal full-length photo the shoes occupy a tiny fraction of the frame, which produces a description like "black low top sneaker" that matches thousands of listings.
Bags have the same problem, made worse by usually being turned away from the camera. Jewellery and eyewear frequently do not register at all from a distance.
All of these do far better from a tight crop of their own. There are separate posts on shoes and bags, because both have their own angles and details that matter.
What does the result tell you about your photo?
The output tells you about your photo, not just about the clothes, once you know what to look for.
Only two items from a full outfit means the framing was too wide or the subject too small.
A garment in an odd category, such as a jacket filed as a Top, usually means the length and the closure were not visible.
A vague colour like "dark" means the light took the colour out. Find a brighter frame.
A thin details field means the model could see a shape but not the construction. Crop tighter.
The wrong person's outfit means a group shot. Crop to the person you meant.
None of these are the search failing. They are an accurate report of what the picture contained, which is genuinely useful if you read it that way rather than searching again with the same input.
What habit makes this easier?
Screenshot first, decide later.
The hardest part of this is not the search, it is finding the source again. A Story is gone in a day. A Live cannot be rewound. A video you scrolled past is difficult to surface a second time.
So capture it the moment you see it, before you have worked out whether you care. Half a second of effort, and your camera roll becomes a queue you can work through whenever you have a spare five minutes.
Every search you run is saved as well, so nothing is lost by searching something you are unsure about. You can come back to it weeks later, and re-running the shopping search then looks at a genuinely different set of listings, which is worth doing for anything that was sold out the first time.
What if it still comes back empty?
Try a different frame of the same outfit before you give up on it. A second screenshot, one taken a moment earlier or later, is usually enough. The detection is doing the same work each time, so a clearer picture is the one variable actually worth changing.
If the pieces were found but the shopping results look similar rather than identical, that is a different situation with its own reasons. There is a post about why that happens.