Wan 2.6 API

Wan 2.6 Text to video turns a written shot, or a numbered list of shots, into a clip of exactly 5, 10 or 15 seconds at 720p or 1080p. Only prompt is required. Everything else has a default: 16:9, 720p, five seconds, one continuous take and prompt expansion off. The clip comes back with sound generated alongside the picture, and audio_url can hand it a recorded line or music cue to follow instead.

Choose a Wan 2.6 mode by starting material

Five shapes at both resolutions

aspect_ratio accepts 16:9, 9:16, 1:1, 4:3 and 3:4, and every shape is available at 720p and at 1080p. Nothing is inherited from an input frame, so the same brief can be rendered as a landscape master and as a vertical cut-down by changing one field. Resolution sets the per-second rate and how long the task takes, which makes the default 720p the natural place to settle a prompt.

One take or a planned sequence

shot_type switches between single, one continuous take, and multi, where the model may cut between several shots inside the same clip. A multi request always runs with prompt expansion on, whatever enable_prompt_expansion says, because the cuts are planned while the brief is expanded. Multi suits a short story with a beat to it; single is the choice when continuity matters more than coverage.

Sound in the same pass

Audio is generated with the picture, so a sentence about sound at the end of the prompt shapes what you hear. When the sound already exists, audio_url takes a WAV or MP3 of 3 to 30 seconds and up to 15 MB, and mouth shapes and movement pace themselves to that file. A longer track is cut to the clip length; with a shorter one, the rest of the clip is silent.

An atmospheric single shot with written sound

The published harbour example shows the single-take pattern. A fisherman in a yellow oilskin coils a wet rope, looks up at a passing ferry and raises a hand, seen through one slow dolly from behind a stack of crates. The prompt closes with an audio line, so steady rain, a long ferry horn and the creak of the rope sit under the picture without a separate track. It runs at 16:9, five seconds, 720p.

A vertical micro-story for social feeds

The bakery example asks for three numbered shots in a 9:16 clip of ten seconds at 1080p: the shutter rolling up at dawn, floured hands shaping a loaf, then the oven door glowing as a tray goes in. shot_type is multi, so each numbered line gets its own cut, and a fixed seed keeps the draw steady while the wording is refined.

Look development before anything is shot

A written brief is enough to test setting, light and pacing for a scene that does not exist yet. It will not hold a specific face or product between runs. Once a person or a package has to stay recognisable, create the frame first and move the job to image-to-video, where the opening image fixes the appearance.

# 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-2.6/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "duration_seconds": 5,
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
    "resolution": "720p",
    "shot_type": "single"
  }
}'

# 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.