Model · Meta Muse

Muse Video API

Generate 5 to 60 second videos with sound from Meta's Muse model through one async endpoint, POST /v1/videos. Meta's developer platform has no video model yet. On LaoZhang API a clip costs $0.30 to $1.55, and you don't need a Meta account.

Prices and availability checked November 9, 2026

Get API Key Watch sample clips Read the docs

per clip of 5, 8 or 10 seconds: $0.30
longest single clip, $1.55, with sound: 60 s
async: create a task, poll, download: /v1/videos
one LaoZhang key for every model: No Meta account

Is there an official Muse video API?

No. On November 9, 2026, Meta's developer platform (Meta Model API) offered Muse Spark, Muse Wonder, Muse Image and other models, but no video model. People make Muse videos by hand in the Muse app. Developers who need them in code use an independent integration such as muse-video on LaoZhang API.

Ways to get Muse video, checked November 9, 2026
Option	API access	How it's paid	Notes
Meta Model API (dev.meta.ai)	No video model	Not applicable	Text, image and speech models only
Muse app	None; videos are made in the app	Power $20 a month, Max $100 a month	Plans are set in weekly tokens, and Meta doesn't say how many videos they cover. Muse is in limited testing and not offered in every country.
LaoZhang API, muse-video	POST /v1/videos with an API key	$0.30–$1.55 per clip, prepaid	No Meta account. The same key calls GPT, Claude, Gemini and other models.

With no public Meta video API, there is no official per-video price to compare against. The Muse app plans above are Meta's only published prices for Muse.

LaoZhang API isn't affiliated with Meta. muse-video is an independent integration, not a Meta product, so Meta can change or limit it without notice. If the route pauses, the status page and docs say so, and tasks that fail are refunded in full.

Price per clip

Each model makes one continuous clip of a fixed length and charges one price. Aspect ratio, reference images and, for muse-video, the choice of 5, 8 or 10 seconds don't change it.

Muse models on LaoZhang API
Model ID	Clip length	Per clip	Per second at full length
muse-video	5, 8 or 10 s	$0.30	$0.030
muse-video-20s	20 s	$0.55	$0.028
muse-video-30s	30 s	$0.80	$0.027
muse-video-40s	40 s	$1.05	$0.026
muse-video-50s	50 s	$1.30	$0.026
muse-video-60s	60 s	$1.55	$0.026

You're charged when the task is created. Failed tasks are refunded in full, and requests rejected at submission with a parameter error (HTTP 400) aren't charged. Every resubmission is a new task and a new charge, so keep the task id and poll it instead of sending the request again.

Any default, vip or svip token with Usage first or Per-call billing can call Muse; all three groups pay the same price. The usage log shows the charge for each task.

Sample clips

Unedited Muse output with the soundtrack the model generated. Click a clip to play it; each caption gives the model, length, aspect ratio and prompt.

Our November 9, 2026 test: muse-video, 10 s, 16:9. Prompt: A night market street in the rain, steam rising from a noodle stall, a cook tossing noodles in a wok, neon signs reflecting in puddles, slow dolly forward, sizzling wok and rain sounds

muse-video, 10 s, 16:9, style_preset 2d_anime with a custom camera move. Prompt: A girl with a red scarf runs across a rooftop at sunset as birds fly past

muse-video-40s, 40 s, 16:9, one continuous clip. Prompt: A hot air balloon rises over a valley at dawn, drifts over a river and a village, then lands in a meadow

muse-video-50s, 50 s, 9:16 portrait. Prompt: A cyclist rides through a coastal town from morning to sunset: market, harbor, cliff road, lighthouse

More samples, including continued clips

Our test call, November 9, 2026

One muse-video request with the night market prompt above, sent with an ordinary default-group token and timed from submission to completion.

muse-video test result (UTC+8)
Item	Result
Request	muse-video, seconds 10, aspect_ratio 16:9
Time to completed	151 s (created 15:06:02, completed 15:08:33)
Output	1280×720, 24 fps, 10.0 s, H.264 MP4
Audio	AAC mono, 24 kHz, generated with the clip (mean level −15.9 dB)
File size	11.6 MB
Charged	$0.30 in the usage log, at task creation

The task returned on the first attempt. The file carries no content credentials (c2patool finds no C2PA manifest) and no metadata that names a generator, so unlike our GPT Image results it shows what the route returns, not where it was rendered. The clip is the first one under Sample clips.

Make your first clip

1Create a key
Register and create a token in the default group with Usage first billing. No Meta account or video group is needed.
2Create a task
POST /v1/videos with model and prompt. It returns a task id within seconds; save it.
3Poll the task
GET /v1/videos/{id} every 15 seconds. 10-second clips usually finish in 2–3 minutes, 40–60 second clips in about 5–10. Keep polling for up to 30 minutes.
4Download
When status is completed, download the MP4 from video_url. Links last 7 days, so save the file.

curl (needs jq)

export LAOZHANG_API_KEY="YOUR_LAOZHANG_API_KEY"

# 1. Create the task
VIDEO_ID=$(curl -s https://api.laozhang.ai/v1/videos \
  -H "Authorization: Bearer $LAOZHANG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "muse-video",
    "prompt": "A lighthouse on a rocky coast at dawn, waves crashing, slow push-in",
    "seconds": "10",
    "aspect_ratio": "16:9"
  }' | jq -r .id)

# 2. Poll until completed or failed
while true; do
  TASK=$(curl -s "https://api.laozhang.ai/v1/videos/$VIDEO_ID" \
    -H "Authorization: Bearer $LAOZHANG_API_KEY")
  STATUS=$(echo "$TASK" | jq -r .status)
  echo "$STATUS"
  [ "$STATUS" = completed ] || [ "$STATUS" = failed ] && break
  sleep 15
done

# 3. Download
curl -L -o muse-video.mp4 "$(echo "$TASK" | jq -r .video_url)"

Python (requests)

import os
import time

import requests

BASE_URL = "https://api.laozhang.ai"
HEADERS = {"Authorization": f"Bearer {os.environ['LAOZHANG_API_KEY']}"}

task = requests.post(
    f"{BASE_URL}/v1/videos",
    headers=HEADERS,
    json={
        "model": "muse-video-30s",
        "prompt": "A paper boat drifts down a rainy city gutter, then sails into a sunny pond",
        "aspect_ratio": "9:16",
    },
    timeout=120,
)
task.raise_for_status()
video_id = task.json()["id"]

deadline = time.time() + 30 * 60
while True:
    video = requests.get(f"{BASE_URL}/v1/videos/{video_id}", headers=HEADERS, timeout=60).json()
    if video["status"] == "completed":
        break
    if video["status"] == "failed":
        raise RuntimeError(video["error"]["message"])
    if time.time() > deadline:
        raise TimeoutError(f"{video_id} is still running; poll the same id later")
    time.sleep(15)

with open("muse-video.mp4", "wb") as f:
    f.write(requests.get(video["video_url"], timeout=300).content)

Send a first frame in image, a closing frame in last_frame, or up to 4 reference_images. The 20s to 60s models also take video_url to continue an existing clip, and style_preset or skill_preset to pick one of 7 built-in looks or 13 camera techniques.

All request fields and errors

Limits to check before you build on it

1720p at 24 fps
Output is 1280×720 landscape or 720×1280 portrait (960×960 square). Sending 1080p is accepted but doesn't change the size. For 1080p use Seedance or Wan; for 2K, MiniMax H3.
2Faces aren't locked
Reference images carry over clothing, hair and style, not the same face. Compare other video models if the same face across clips matters.
3Sound is always on
Every clip has music and effects shaped by the prompt. There's no switch to turn it off; strip the audio track after download if you don't need it.
4Async and slow at peak
Busy periods can push a 60-second clip close to 20 minutes. Don't resubmit while a task is running: it charges again. A task still unfinished after 60 minutes is marked failed and refunded.
5Not a Meta API
The route is an independent integration and can change if Meta changes Muse. Muse video files from it carry no content credentials, so a file can't prove where it was made.
6Data and service terms
LaoZhang API doesn't store prompt or response content by default. Video links expire after 7 days. The service is provided "as is" and "as available"; the status page shows current availability.

Frequently asked questions

Does Meta have an official Muse video API?

Not as of November 9, 2026. Meta Model API offers Muse Spark, Muse Wonder, Muse Image and other models but no video model, and the Muse app has no API. LaoZhang API's muse-video is an independent integration, not affiliated with Meta.

How much does a Muse video cost through the API?

$0.30 for a 5, 8 or 10 second clip with muse-video, rising by $0.25 per 10 seconds to $1.55 for a 60-second clip with muse-video-60s. Aspect ratio and reference images don't change the price. Failed tasks are refunded.

Do I need a Meta account or a Muse subscription?

No. You need only a LaoZhang API key from a default, vip or svip token with Usage first or Per-call billing. The same key works for the other models on LaoZhang API.

How long does a Muse video take?

Clips up to 10 seconds usually finish in 2–3 minutes, 20–30 second clips in about 4–10, and 40–60 second clips in about 5–10; at busy times a 60-second clip can take close to 20 minutes. Poll the task every 15 seconds for up to 30 minutes.

Can I get 1080p or turn the sound off?

No. Muse outputs 720p-class sizes at 24 fps, and every clip has a soundtrack. Remove the audio after download if needed, or use Seedance or Wan for 1080p.

Can I animate an image or extend a clip?

Yes. Send image for the first frame, last_frame for the closing frame, or up to 4 reference_images. To extend a clip, send video_url to a 20s–60s model; the finished video is as long as the model name says, source included.

What happens if Meta changes Muse?

The route can pause or change without notice because it isn't a Meta API. The status page and docs announce changes, and any task that fails is refunded in full. Keep a second video model in mind for production work.