Introduction
Wan 3.0 Prime Text to video generates a 2 to 30 second clip with sound from a written prompt alone. Its fields match Wan 3.0 Text to video exactly: prompt is required, while duration_seconds, resolution (480p, 720p or 1080p, default 720p), aspect_ratio (default 16:9), generate_audio, enable_prompt_expansion and seed are optional. An existing Wan 3.0 request therefore works here with only the model name changed. Alibaba positions Prime as the high-speed version of Wan 3.0 with the same capabilities.
Choose a Wan 3.0 Prime mode by starting material
| What you have at the start | Mode |
|---|---|
| a written scene or script | Wan 3.0 Prime Text to video |
| an approved opening image | Wan 3.0 Prime Image to video |
| separate visual, motion, or sound references | Wan 3.0 Prime Reference to video |
Key features & highlights
A drop-in request with faster turnaround
Because the schema, defaults and enums are identical, moving a job between Wan 3.0 and Prime is a one-word change. Send deadline work here, and keep jobs on Wan 3.0 when per-second price matters more than turnaround. Holding the seed on both models lets you compare the two files directly.
Long takes you can check for continuity
The lighthouse example runs 10 seconds at 720p with audio: a revolving beam sweeps across two people every few seconds and leaves them in the dark between passes, with a foghorn and the sea on the soundtrack. A periodic subject is an honest test of whether a long clip stays continuous or quietly resets.
Backlight, sheer fabric and wind
The rooftop example is five seconds at 720p: a sheer linen shirt lifting behind a woman on a Mediterranean ledge at golden hour, a slow orbit, warm backlight and lens bloom, sea breeze and gulls. Backlit sheer fabric and moving hair are demanding subjects, so they are worth checking closely in your own renders.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | The shot itself: subject, one action, one camera move, then light and sound. Long lists of adjectives move the result less than one clear action does. at least 1 characters |
aspect_ratio | No | 16:9, 9:16, 1:1, 4:3, 3:4 or adaptive, which lets the model choose. Defaults to 16:9. |
enable_thinking | No | Has the model think through the brief and any reference material before generating. It helps with dense documents, web pages and complex reference images, and makes the run slower. Default: |
resolution | No | 480p, 720p or 1080p. Defaults to 720p. |
duration_seconds | No | Any whole number from 2 to 30, or -1 for smart duration, where the model picks the length: the 30-second maximum is held up front and the charge settles on the length actually d… |
generate_audio | No | On by default. The clip comes back with sound already locked to the picture. Default: |
enable_prompt_expansion | No | Off by default. When on, the prompt is rewritten before generation, which helps thin briefs and fights you on detailed ones. Default: |
seed | No | 0 to 2147483647. Hold it steady to compare one changed field at a time; drop it for variety. |
Pricing
| Specification | Price |
|---|---|
| $0.0612 / second |
| $0.126 / second |
| $0.252 / second |
Use cases
Surreal set pieces that must stay consistent
The salt-flat example asks for a mirror-flat flood where sky and ground meet with no visible horizon, a woman walking toward the camera while her reflection walks upside down beneath her, and rings spreading across the water when she spins. The frame starts locked and only pushes in afterwards, which isolates the reflection as the thing to review.
Performance with the voice in the same pass
The recording-booth example has a singer holding a long low note at a ribbon microphone, pulling one ear cup off and smiling through the control-room glass, with a rack focus from the microphone mesh to her face written into the prompt. The voice is generated in the same pass as the picture, so both come back in one file.
Prompt guide & pro tips
Write camera changes as events
When the camera does something mid-shot, write it as an event in sequence: the booth prompt says "rack focus from the microphone mesh to her face", and the salt-flat prompt says "wide static frame that then pushes slowly in". One camera instruction per beat keeps the result easy to review.
Give longer clips a rhythm
For clips longer than 5 seconds, give the render something that repeats or progresses — a beam every few seconds, a tide, a room filling up — so continuity can be judged. Durations are any whole number from 2 to 30, or -1 to let the model choose; settle the look at 5 seconds first.
Compare Prime and Wan 3.0 with the seed held
Send one body to both models with the same seed, then change one field at a time. A fixed seed makes the comparison fair but does not guarantee identical frames. Leave enable_prompt_expansion at its default false for detailed prompts and turn it on only for a thin two-line idea.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${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: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "alibaba/wan-3.0-prime/text-to-video",
"input": {
"prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
"resolution": "720p",
"aspect_ratio": "16:9",
"duration_seconds": 5
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.
