Wandura C1.1: ad videos from a description

TL;DR: POST /v2/jobs with model: "wandura-c1-1", a prompt that describes the ad, and up to 4 reference images. You get back one finished video of 5 to 60 seconds. Wandura plans the shots, renders them, and joins them. You do not write a system prompt. The price is per output second: 480p $0.09, 720p $0.15, 1080p $0.30. Poll GET /v2/jobs/{job_id} or receive the signed webhook, exactly as for other video jobs. Failed, blocked and timed-out jobs are free.

Basics#

• Model id: wandura-c1-1 ("Wandura C1.1"), output video. It is called through the Jobs API with the same keys, idempotency, webhooks and statuses as other jobs.
• No style. Send the description and the settings as top-level fields. A style or an inputs object is refused.
• What it makes: one continuous ad video. The video keeps the people, products and places in your reference images.

Request#

curl -X POST https://api.wandura.ai/v2/jobs \
  -H "Authorization: Bearer $WANDURA_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ad-2026-10-20-001" \
  -d '{
    "model": "wandura-c1-1",
    "prompt": "A 15 second ad for Aurora, a meditation app. The woman in Image 1 sits by a window at sunrise, closes her eyes and breathes; the phone in Image 2 shows the app. Calm piano. She says: \"Ten minutes changed my mornings.\" End on the app logo.",
    "image_urls": ["https://your-bucket.example.com/ads/woman.jpg", "https://your-bucket.example.com/ads/phone.png"],
    "duration": 15,
    "resolution": "720p",
    "aspect_ratio": "9:16",
    "webhook_url": "https://your-backend.example.com/hooks/wandura",
    "end_user_id": "ua_member_17"
  }'
# HTTP 202
# {"job_id":"run_7Hq2LmPaX9Ws","status":"queued","model":"wandura-c1-1","created_at":"2026-10-20T02:14:46.251Z"}

• Input URLs are downloaded when the job runs, not when you submit. Keep presigned URLs valid for at least 30 minutes.
• Send an Idempotency-Key header on every submit so a retry never makes a second video (see Idempotency).

Poll and output#

curl https://api.wandura.ai/v2/jobs/run_7Hq2LmPaX9Ws -H "Authorization: Bearer $WANDURA_KEY"

{"job_id":"run_7Hq2LmPaX9Ws","status":"succeeded","model":"wandura-c1-1","style":"sty_wandura_c1_1",
 "output":[{"url":"https://<wandura-storage-host>/outputs/2026-10-20/<random name>.mp4","type":"video/mp4",
   "width":720,"height":1280,"duration":15,"fps":30,"size_bytes":4183022,"seed":null,"expires_at":"2026-10-27T02:22:21.000Z"}],
 "error":null,"created_at":"2026-10-20T02:14:46.251Z","completed_at":"2026-10-20T02:22:21.964Z"}

• The body has the same fields as every other video job, so one parser handles all of them. For this model, style is a fixed value; ignore it.
• Statuses: queued → processing → succeeded | failed | cancelled | timeout (see Statuses).
• How long it takes: 2 to 8 minutes in our tests (5 to 30 s videos). Longer videos are rendered in parallel parts, so a 60 s video does not take 12 times as long as a 5 s one. Poll every 10 to 15 seconds, or use the webhook.
• Output: one MP4 on Wandura storage. The link works for at least 7 days (expires_at). Copy the file if you need it longer (see Output).

Webhook#

With webhook_url, Wandura POSTs the final job body (the same JSON as the poll) once the job ends, signed with X-Signature and X-Timestamp. Verification, retries and deduplication work exactly as for other jobs: see Webhooks and Verify the signature.

Prices#

The price is per output second, by resolution. It covers the whole job: planning, rendering and joining.

• When the job is accepted, the price for the duration you asked for is held on your wallet. When it succeeds, you are charged for the seconds delivered, never more than the hold.
• failed, timeout and cancelled jobs and refused submits (4xx) are free. The hold is released.
• The relax lane costs the same as priority for this model.
• Read the live prices from GET /v2/models/wandura-c1-1 (price_per_second) or GET /v2/public/workflows.
• Test keys (wdr_test_…) are free: the job succeeds at once with a sample video and nothing is held.

Limits#

When many C1.1 jobs run at once, new ones wait as queued until there is room. They are not refused.

Errors#

A bad body is refused at once with 422 INVALID_INPUT, before any money is held. Every problem is listed:

HTTP 422
{"error":{"code":"INVALID_INPUT","message":"Validation failed.","retryable":false,"details":[
  {"field":"image_urls","reason":"at most 4 reference images"},
  {"field":"duration","reason":"must be whole seconds from 5 to 60"},
  {"field":"style","reason":"wandura-c1-1 takes no style"}],
 "docs_url":"https://docs.wandura.ai/developers/api-v2/errors#jobs-invalid-input"}}

When your product has reached its daily spend cap, the submit is refused and nothing is held:

HTTP 429
{"error":{"code":"DAILY_CAP_REACHED","message":"This product reached its daily spend cap (the day resets at 00:00 Vietnam time, UTC+7). Ask your org admin to raise the cap.","retryable":false,"docs_url":"https://docs.wandura.ai/developers/api-v2/errors#jobs-daily-cap-reached"}}

A job that ends without a video is not an HTTP error. Its error.code is GENERATION_FAILED (retryable), CONTENT_MODERATION_BLOCKED (change the description or images) or TIMEOUT (submit again). All three are free. See Statuses.

Remake a source ad#

Send an existing ad as video_url and get a new ad with the same scenes and pacing, with your people and your product.

curl -X POST https://api.wandura.ai/v2/jobs \
  -H "Authorization: Bearer $WANDURA_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: remake-2026-10-20-001" \
  -d '{
    "model": "wandura-c1-1",
    "video_url": "https://your-bucket.example.com/ads/source-ad.mp4",
    "prompt": "Remake it for Lumi sparkling water: a summer sale at a Lumi drinks store. The woman in Image 1 is the shop assistant.",
    "image_urls": ["https://your-bucket.example.com/ads/woman.jpg"],
    "resolution": "720p",
    "aspect_ratio": "16:9"
  }'
# HTTP 202
# {"job_id":"run_Pz81KqTmV2Rc","status":"queued","model":"wandura-c1-1","created_at":"2026-10-20T05:14:46.251Z"}

• What stays: the scenes in order, with their timing scaled to the new length, the shots and camera moves, the pacing, the music and sound, and the shape of the lines and on-screen text. What changes: the people (yours from the images, or new ones), the product and brand (yours from the prompt, or a generic one), and the lines, rewritten for your product. Real brands, logos and people from the source ad are not copied.
• Read once: Wandura reads a source ad once and reuses that reading for every later remake of the same video file, from any link or a re-upload. A Wandura C1.1 remake sent back as video_url reuses the reading of its original source ad. The reading only describes the source ad. It never holds your prompt, images or brand.
• Price: the per-second prices above. Reading the source ad costs nothing extra. The hold is for duration, so for the source ad's length when you do not send one.
• Checked before any hold: a video_url that is not a readable MP4 or MOV, or that is longer than 60 s, gets 422 INVALID_INPUT and nothing is held. Keep the link valid for at least 30 minutes, because the file is downloaded again when the job runs.
• Maximum job time: 1080 s (18 minutes) from submit on the priority lane, the same as above.
• Ends without a video, free: a source ad with adult, violent, hateful or dangerous content ends failed with CONTENT_MODERATION_BLOCKED. A link that no longer serves a video of at most 60 s when the job runs ends with INPUT_UNUSABLE. A source host that is busy ends with GENERATION_FAILED (retryable).

HTTP 422
{"error":{"code":"INVALID_INPUT","message":"Validation failed.","retryable":false,"details":[
  {"field":"video_url","reason":"the source ad lasts 75 s; at most 60 s"}],
 "docs_url":"https://docs.wandura.ai/developers/api-v2/errors#jobs-invalid-input"}}