Animate a child's drawing with AI: 6 seconds for $0.075 on Sume

The short answer

To animate a child's drawing, photograph it flat, put the image at a public HTTPS URL, and send it as the first frame to grok-imagine-video-1.5 on Sume. A 6-second clip at 720p costs $0.075 (6 seconds at $0.0125). The model is image-to-video only, takes 4 to 15 seconds, and returns no audio.

Why Grok Imagine suits a drawing

Grok Imagine Video 1.5 is the cheapest image-to-video row in the Sume catalog, which makes it a good fit when you will make many small clips: one per drawing in a class set, a birthday card, or a school newsletter. Sume bills the provider list of $0.01 per output second times the 1.25 house margin, so every second is $0.0125 at 480p or 720p.

What each length costs

Duration is the only thing that changes the price. The catalog row allows 4 to 15 seconds.

The request

Send the drawing as a first_frame in frame_images. Grok needs an image: the catalog row says image_url or first_frame_url is required, so a text-only request fails. There is no aspect_ratio, generate_audio, or bitrate_mode field on this model, so leave them out.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: drawing-001" \
  -d '{
    "model": "grok-imagine-video-1.5",
    "prompt": "The crayon dragon flaps its wings and the sun sparkles",
    "duration": 6,
    "resolution": "720p",
    "frame_images": [{
      "type": "image_url",
      "image_url": {"url": "https://example.com/dragon.png"},
      "frame_type": "first_frame"
    }]
  }'

Poll and download

The job is asynchronous. The submit call returns 202 with an id and a polling_url; poll it until status is completed, then download from unsigned_urls[0] or GET /v1/videos/{id}/content. Send an Idempotency-Key so a retry cannot bill twice.

Prepare the picture, and when to pay more

Photograph the page straight on, in even light, with the drawing filling the frame. The first frame is what the model animates, so a skewed phone photo with a thumb in the corner becomes a moving skewed photo. Work from the child's artwork, not from a photo of the child.

Grok has no last frame. If you want the drawing to end on a specific picture, such as the same dragon colored in, use kling-3 with a last_frame. Kling 3 is $0.14 a second silent, so 5 seconds is $0.70 and 4 seconds is $0.56.

Add sound afterwards

Grok returns a silent clip. If the clip should have a song or a voice, add it in a Timeline render with the soundtrack field, which costs $0.10 per output minute rounded up. For a class set, loop the same request over each drawing with its own idempotency key, so one failed job does not repeat the other 29.