Skip to main content
Two API calls and you’re done: upload the customer photo, run the try-on. This guide shows the recommended pattern — pre-upload the customer image, ship the product inline. You don’t need a separate “create product” step.

​
Prerequisites

  • An API key ( create account on app.genlook.app )
  • At least 1 credit on your account
  • A product image (URL or local file)
  • A customer photo (JPEG, PNG, WebP, or HEIC, max 10 MB)

​
Step 1 — Upload the customer photo

curl -X POST "https://api.genlook.app/tryon/v1/images/upload" \
  -H "x-api-key: gk_your_api_key" \
  -F "file=@customer.jpg"
Pass crop=false in the form data if you want to keep the original framing (default is a 4:5 person-aware crop). You get back an imageId. Reuse it across as many try-ons as you want against the same photo.

​
Step 2 — Run the try-on

Reference an existing product by externalId, or upsert one inline. The inline form is great for first-time products; the reference form is the cheap repeat-call shape. 
curl -X POST "https://api.genlook.app/tryon/v1/try-on" \
  -H "x-api-key: gk_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "products": [{
      "externalId": "shirt-42",
      "title": "Red tee",
      "description": "Soft cotton regular fit",
      "images": [{ "source": { "url": "https://cdn.example/red-tee.jpg" } }]
    }],
    "person": { "image": { "source": { "id": "<imageId-from-step-1>" } } }
  }'
The next time you generate against shirt-42, just send { products: [{ externalId: "shirt-42" }], person: { image: { source: { id } } } }. The server already has the product cached.

​
Step 3 — Poll for the result

Python
import time

while True:
    r = requests.get(f"{BASE}/generations/{generation_id}", headers=headers)
    data = r.json()
    if data["status"] == "COMPLETED":
        print("Result:", data["resultImageUrl"])
        break
    if data["status"] == "FAILED":
        raise RuntimeError(data.get("errorMessage"))
    time.sleep(2)
The resultImageUrl is a temporary URL — download or display it promptly. You can re-fetch the generation later for a fresh URL.

​
What you skipped

You went straight from “I have an image” to “I have a try-on result” — no POST /products call, no sync loop, no per-product housekeeping. Inline upsert is the default path for a reason:
  • No separate “create product” step. The inline products[] field in /try-on handles that. The dedicated POST /products endpoint is an opt-in alternative for catalog-management cases (see Upsert Product).
  • No catalog sync, no TTL anxiety. Products created inline keep refreshing for 15 days from their last use — keep using them and they stay alive forever.
  • No image-bytes-on-every-call. URLs are cached server-side; uploaded person photos live as imageId for reuse.

​
What’s next

  • TypeScript? — the @genlook/api SDK wraps all three steps in typed methods with automatic retries and a one-line poller, and the Next.js example app is a clonable working integration.
  • Recommended workflow — see Full Example for the ref-first / upsert-on-change pattern, error handling, and what to do when a product expires.
  • All endpoints — see the API Reference (left nav).
  • Try-On options — pre-uploaded vs URL vs multipart customer images, crop control, role hints, anonymous one-shots, all on POST /try-on.