Image

Virtual try-on

Put a garment on a photo of a person and get back one still image. It is the same POST /v1/images endpoint as image edit, with the photos passed in input_references (URL or base64 data URL) — but here the order of the references matters:

• The first reference is always the person photo; every later reference is a garment image.
• At least 2 references are required — fewer returns a 400. Models that take a single garment also return a 400 for more than 2 references.
• prompt is optional for every try-on model. Where the upstream model has a text field it is forwarded; where it has none, a prompt you send is ignored.
• Billed per image (some rows scale with input/output megapixels). The exact charge, platform fee included, is in the response usage.cost; current prices are on the models catalog.

import requests

resp = requests.post(
    "https://videorouter.sh/api/v1/images",
    headers={"Authorization": "Bearer llmr_sk_live_..."},
    json={
        "model": "fal/virtual-try-on-v2",
        # person photo FIRST, then the garment image(s)
        "input_references": [
            {"type": "image_url", "image_url": {"url": "https://example.com/person.jpg"}},
            {"type": "image_url", "image_url": {"url": "https://example.com/jacket.jpg"}},
        ],
    },
).json()

const res = await fetch("https://videorouter.sh/api/v1/images", {
  method: "POST",
  headers: { Authorization: "Bearer llmr_sk_live_...", "Content-Type": "application/json" },
  body: JSON.stringify({
    model: "fal/virtual-try-on-v2",
    // person photo FIRST, then the garment image(s)
    input_references: [
      { type: "image_url", image_url: { url: "https://example.com/person.jpg" } },
      { type: "image_url", image_url: { url: "https://example.com/jacket.jpg" } },
    ],
  }),
});
const data = await res.json();

curl https://videorouter.sh/api/v1/images \
  -H "Authorization: Bearer llmr_sk_live_..." \
  -H "content-type: application/json" \
  -d '{
    "model": "fal/virtual-try-on-v2",
    "input_references": [
      {"type": "image_url", "image_url": {"url": "https://example.com/person.jpg"}},
      {"type": "image_url", "image_url": {"url": "https://example.com/jacket.jpg"}}
    ]
  }'
# -> {"data": [{"url": "https://..."}], "usage": {"cost": 0.0416, ...}}

Which models

Ten models across four providers. They all share the request shape above; they differ in how many garment images they accept:

Every model here also shows up under the Virtual Try-On filter on the models catalog. Multi-host models fail over to the next-cheapest host automatically — see Provider selection.

Not yet supported

There are no garment-category, pose-preservation or mask parameters — the model infers the fit from the garment photo. stream: true is rejected outright, same as generation.