Video

Videos take minutes, so this works like OpenAI’s video API: create a job, check on it, then download the result.

1. Start a video

POST /v1/videos

curl http://localhost:7860/v1/videos \
  -H "Content-Type: application/json" \
  -d '{"prompt": "a paper boat drifting down a rainy street", "seconds": "4", "size": "832x480"}'

{ "id": "...", "object": "video", "model": "wan2-1-t2v-1-3b-diffusers", "status": "queued", "progress": 0, "seconds": "4.1", "size": "832x480" }

Field	Notes
prompt	Required.
model	auto, a model id or a Hugging Face repo id.
seconds	Length of the clip.
size	"WIDTHxHEIGHT". Leave out for the model’s usual size.
negative_prompt, seed, steps	Optional extras.

To animate a picture, send a multipart form instead, with the picture as input_reference:

curl http://localhost:7860/v1/videos -F prompt="the waves start rolling" -F input_reference=@beach.jpg

2. Wait for it

GET /v1/videos/{id}

status goes queued → in_progress → completed (or failed, with an error). progress is 0 to 100.

3. Download it

GET /v1/videos/{id}/content

Returns the MP4 file.

import time, httpx

base = "http://localhost:7860/v1"
job = httpx.post(f"{base}/videos", json={"prompt": "a paper boat drifting down a rainy street", "seconds": "4"}).json()
while job["status"] in ("queued", "in_progress"):
    time.sleep(5)
    job = httpx.get(f"{base}/videos/{job['id']}").json()
    print(job["status"], job["progress"])
if job["status"] == "completed":
    open("boat.mp4", "wb").write(httpx.get(f"{base}/videos/{job['id']}/content").content)