How virtual try-on works (photo, live, and what it still cannot do)

Virtual try-on is a short name for three different machines. They answer different questions. Mixing them up is why demos look magical in a trailer and disappointing on your own hoodie.

Three generations, one crowded label

| Kind | What you give | What you get | Move? | | -------------------- | ----------------------------- | ---------------- | ----- | | Static overlay | Measurements or a mannequin | Flat collage | No | | Photo generation | A selfie + a garment image | One still result | No | | Live try-on | Camera stream + garment image | Continuous video | Yes |

TryOnAnytime ships photo and live in the studio. We do not pretend a still frame is a tape measure.

What the model is actually doing

Rough pipeline (photo path):

1. Read your body silhouette and pose from the person image.
2. Read the garment region from the product shot.
3. Warp / generate the cloth onto that pose with lighting guesses.
4. Return a single image you can save or compare.

Live path keeps that loop running on camera frames so hems can swing when you turn. That is useful for “does this silhouette fight my shoulders?” It is still not fabric lab data.

What it is good for

• Colour and proportion gut-checks before you buy.
• Comparing two SKUs on the same photo.
• Giving a storefront shopper a preview without leaving the product page (embed).

What it is bad for

• Exact size prediction (XS vs M).
• Sheer, sequins, tiny logos, multi-garment flat lays — see where try-on struggles.
• Group photos or heavy crop that hides the torso.

Photo vs live (pick by job)

Quiet evening, three candidates from one boutique → photo.
Desktop browsing across random shops → live or the any-store / extension flow.

More detail: photo vs live virtual try-on.

Credits and honesty

Runs consume credits (pricing). We will not invent “92% accuracy” scores. If the sleeve looks wrong, trust that miss and open the size guide.

Start here: Open the studio.