Wan 2.5 API

Wan 2.5 Image to video turns the still you send in image_url into a 5- or 10-second clip with sound. The image decides subject, wardrobe, set, light direction and the clip's proportions; the optional prompt, up to 1,500 characters, describes what happens after it. There is no aspect-ratio field on this endpoint, so crop the image to the delivery shape before sending it. With only a written idea and no still, use the text-led endpoint instead.

Choose a Wan 2.5 mode by starting material

The opening frame carries the look

image_url accepts JPEG, PNG, BMP or WebP up to 20 MB, with each side between 240 and 8,000 pixels. Anything already visible in that file, such as a label, a silhouette or a costume, does not need describing again, so the prompt can spend its words on motion, camera and sound.

Audio generated with the picture, or timed to your track

Every clip returns with sound generated alongside the picture. Add audio_url, a WAV or MP3 of 3 to 30 seconds and up to 15 MB, when a recorded line or music cue should set the timing: the face comes from the frame and the mouth timing from the audio. A track longer than the clip is cut at the end; a shorter one leaves the remainder silent.

Two lengths and three resolutions

duration_seconds is exactly 5 or 10, and resolution is 480p, 720p (the default) or 1080p. Higher tiers cost more per second and take longer to return. Short clips hold hands and hair better, so settle the motion at five seconds before committing to ten.

Product motion from one packshot

The published serum-bottle example sends a product still and asks only for a slow quarter-turn camera arc while a band of window light slides across the glass; the bottle itself does not move. Because the frame already fixes label and silhouette, the prompt carries only light and motion, rendered at 720p for five seconds.

A portrait that speaks a recorded line

Send an approved portrait as image_url and the voice line as audio_url. The person's appearance comes from the image and the mouth timing from the file, so the review is about delivery rather than lip sync.

A longer studio take with guarded details

The slip-dress example runs ten seconds at 1080p from a studio portrait: a weight shift, a turn toward the lens and two slow steps under a locked camera. Its negative prompt, "extra people, text, jewellery changes", stops accessories from changing and extra figures from appearing over the longer duration.

Rain, water and other sound with a visible source

In the umbrella example, two figures stay nearly still while rain drums on the umbrella above them, so the motion belongs to the water and the sound has a source you can see. When the event is continuous like this, five seconds is usually enough.

# 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-2.5/image-to-video",
  "input": {
    "duration_seconds": 5,
    "image_url": "https://example.com/reference.png",
    "resolution": "720p"
  }
}'

# 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.