Wan 3.0 Pro Prime API

Wan 3.0 Pro Prime Image to video animates an opening frame on the Prime render and delivers it at 1080p, 2K or 4K for 2 to 30 seconds. image_url is required and becomes the exact first frame; the prompt is optional and should describe motion and camera; last_image_url optionally fixes the final frame. The fields match Wan 3.0 Pro Image to video exactly, so a request built for that endpoint runs here with only the model name changed.

Choose a Wan 3.0 Pro Prime mode by starting material

Two-frame instructions such as focus pulls

A focus pull is two frames by nature, which is what last_image_url is for. The bloom example opens sharp on a flower bud and closes on a supplied frame that is sharp on an unbuttoned shirt behind it, 6 seconds at 4k; the prompt only says when the transfer happens. Very shallow depth of field leaves the render little latitude to land on the right plane without a pinned ending.

The source sets the ceiling, not the tier

Resolution governs the render, not the input. The 2k and 4k tiers are enlarged from a native 1080p render, so feed the sharpest still you have: JPEG, PNG without transparency, BMP or WEBP, 240 to 8000 pixels per side, up to 20 MB and no wider than 8:1.

Holds for scored edits

The alley-light-hold example is a 4-second hold at 2k with audio and prompt expansion both off: "Nobody moves. Only the neon tube above her stutters and the rain keeps falling, dragging its reflection across the wet plastic. Locked camera." It is the request shape for a shot that will be scored in the edit instead of carrying its own sound.

Environmental change around a still subject

The neon-alley example opens on a plate that already carries the alley, the rain and the transparent slicker, so the prompt says only what changes: the rain coming down harder, a neon tube stuttering once, steam surging from a vent, and the subject holding the lens through all of it. Its audio line names each source — rain on steel, a buzzing tube, a hissing vent.

Close detail with a fixed end

The shirt example reuses the same two frames as the bloom example at 1080p for 4 seconds: a hand draws an open shirt slowly aside under raking gold light and lets it fall back, while the bud between the folds opens a little. Pinning both ends keeps a very shallow depth of field landing on the intended plane.

# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
  -d '{
  "model": "alibaba/wan-3.0-pro-prime/image-to-video",
  "input": {
    "aspect_ratio": "adaptive",
    "duration_seconds": 5,
    "image_url": "https://example.com/reference.png",
    "resolution": "1080p"
  }
}'

# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $SPICY_API_KEY"

See the API reference for authentication, the complete schema, polling, and error handling.