Generating renders

Two task families on one workflow. The Product beautifier renders the product alone on a background; a photo shoot puts it on a model. Both return a task you poll or subscribe to.

Every task shares the same shape: the same authentication, the same lifecycle queued → processing → completed | failed, the same delivery contract. Pick the endpoint by what you want back.

Product beautifier

A clean render of the product alone on a chosen background. subtype selects the flow.

curl https://api.veeton.com/v1/tasks/beautifier \
  -H "Authorization: Bearer $VEETON_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "subtype": "clothing",
    "inputs": { "product_id": "01HX5K2MZ7A3Q4FBNDC0EVDXY1", "background": "#ffffff",
                "angle_shot": "front", "style": "ghost_mannequin" }
  }'

background is a slug from the backgrounds catalog, shared by every beautifier subtype. Hex-coded slugs include the leading # and are case-insensitive. The full galleries of backgrounds, angle shots and styles are in the API reference.

Photo shoot

The product worn by a model, on a scene of your choice. A photo shoot (tryon) has a single flow: the product’s own type drives the worker, so there is no subtype.

curl https://api.veeton.com/v1/tasks/tryon \
  -H "Authorization: Bearer $VEETON_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "inputs": [
      { "product_id": "01HX5K2MZ7A3Q4FBNDC0EVDXY1",
        "model_id": "01HX5K2MZ7A3Q4FBNDC0EVDXMD1",
        "outfit_product_ids": ["01HX5K2MZ7A3Q4FBNDC0EVDXPN"],
        "background": "park",
        "angle_shot": "full_body_relaxed_symmetric",
        "gaze": "camera" }
    ]
  }'

The body is always an array of 1 to 20 inputs and the response is always { "data": Task[] }, one task per input in submission order.

Consistency groups

Inputs that share the same model_id, product_id and outfit_product_ids are dispatched together as a consistency group: identity and styling are computed once and reused, so the outputs look like one set. Vary angle_shot, background or gaze across the entries of one request to get coherent variants. Separate requests do not share a group.

Validation

If any input fails validation the whole request is refused with 400 and every bad input listed, so you never retry-and-bisect.

{
  "errors": [
    { "input_index": 1, "type": "not_found", "code": "model_not_found", "param": "model_id",
      "message": "Model 01HX… was not found." }
  ]
}

Edit with AI

Change a finished result with a prompt. Name the image to edit and say what you want changed.

{
  "inputs": {
    "image_id": "01HX5K2MZ7A3Q4FBNDC0EVDXIM",
    "prompt": "Remove the wrinkle on the left sleeve."
  }
}

You can edit the output of a completed beautifier or tryon task, or of a completed edit. Your own uploaded photos, and results still running or that failed, are refused with 422 image_not_editable.

An edit makes a new image with its own id. The image you named never changes, and neither does its task’s output_image. Both images are listed in GET /v1/images and in exports, and the new one names its source in edited_from_image_id. To follow a chain of edits, an image nothing was edited from is the end of it. Which version to publish is your choice.

Reading a task

{
  "id": "01HX5K2MZ7A3Q4FBNDC0EVDXY2",
  "type": "tryon",
  "subtype": null,
  "status": "completed",
  "product_id": "01HX5K2MZ7A3Q4FBNDC0EVDXY1",
  "inputs": { "product_id": "01HX5K2MZ7A3Q4FBNDC0EVDXY1", "background": "wood", "angle_shot": "front", "style": "ghost_mannequin" },
  "output_image": { "id": "01HX…", "mime_type": "image/webp", "width": 1536, "height": 2048, "url": "https://…signed…", "url_expires_at": "2026-09-03T15:32:00Z", "created_at": "2026-09-03T14:31:40Z" },
  "failure_reason": null,
  "started_at": "2026-09-03T14:30:12Z",
  "finished_at": "2026-09-03T14:31:40Z",
  "created_at": "2026-09-03T14:30:09Z"
}

inputs echoes what the task was created with, so a task can be understood on its own. output_image is set once the task has completed and is the same Image object every endpoint returns. Its id is an image id: use it with GET /v1/images/{id} for a fresh URL, and it is what the export lists under the product with role: beautifier or tryon.

GET/v1/tasks lists your tasks newest first with status, type, subtype and product_id filters. Use it for reconciliation after webhook downtime; the canonical “tell me when it finishes” channel is a webhook.

Billing

Each task consumes credits from your subscription when it is created. A request without an active subscription or with insufficient credits is refused with 402 payment_required and nothing is queued. A task that fails to dispatch is marked failed and refunded.