Vidu Q3 Pro API

Vidu Q3 Pro is the Pro tier of Vidu's Q3 video line, available as a single image-to-video endpoint. You send an opening frame in image_url and a brief in prompt, both required, and get back 1 to 16 whole seconds of 540p video with audio generated in the same pass. The frame settles subject, look and the shape of the output – there is no aspect ratio field, and the image's own ratio must fall between 1:4 and 4:1 – so the prompt only has to describe what moves.

One resolution, so length is the only size decision

resolution accepts a single value, 540p, and every clip comes back at that size. That leaves duration_seconds – any whole number from 1 to 16, default 5 – as the setting that changes both the length of the clip and the price, and the live estimate follows it as you move it.

Identity comes from the still

Subject, wardrobe, framing and output shape are settled by the image, not by the prompt. Two runs from the same frame differ in movement, not in who is on screen, which is what lets a series of clips from one still cut together. It also means the video request cannot rewrite anything that was approved in the image pass.

Write the motion, then set how much

movement_amplitude sets the overall amount of motion – auto by default, or small, medium, large – while the prompt says what moves and how. When a take moves too much or too little, keep seed and the frame fixed and reword only the movement – a smaller turn, a slower walk – so the difference you see comes from that one edit rather than from a fresh draw.

Reaction sets for one character

One costume frame becomes a wink, a wave, a head tilt and a turn – four requests with the same image and four different prompts. Because identity is fixed by the frame, the set cuts together without anyone drifting between clips.

Close, rhythmic action

The forge example stays tight on a smith's chest and forearms for six seconds, lit from below by the fire, with sparks bursting on each strike of the hammer. Repeated impacts are a useful check for this endpoint: the sparks have to land on the strikes and not between them.

Vertical deliverables

Export the still at 9:16 and the clip comes back vertical. In the underpass example the prompt carries only the walk toward the lens under the strip lights and a camera retreating at her pace; the vertical framing was decided in the image pass.

Uncensored scenes from your own frame

Your still animates as supplied instead of being softened, and because the image carries the content the result stays close to what was approved. Vidu Q3 Spicy in the same line is the LoRA-tuned, uncensored variant.

# 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": "vidu/q3-pro/image-to-video",
  "input": {
    "duration_seconds": 5,
    "image_url": "https://example.com/reference.png",
    "movement_amplitude": "auto",
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
    "resolution": "540p"
  }
}'

# 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.