Video

Video generation

Make videos with Seedance, Omni, Kling, Wan, Veo, Grok, Muse, Vidu, HappyHorse, PixVerse and MiniMax from a prompt, from a first frame (and optionally a last one), or from reference images, videos and audio. Videos take a few minutes, so you start one and check back.

POST/v1/videos
ModelIDPrice per secondStarts fromSoundLength
Seedance 2.5seedance-2.5480p $0.135 · 720p $0.305 · 1080p $0.75Prompt, first and last frame, or up to 9 reference imagesOn by default, free4 to 30 s
Seedance 2.0seedance-2.0480p $0.097 · 720p $0.199 · 1080p $0.445 · 4K $0.99Prompt, first and last frame, or up to 9 reference imagesOn by default, free4 to 15 s
Seedance 2.0 Fastseedance-2.0-fast480p $0.074 · 720p $0.159Prompt, first and last frame, or up to 9 reference imagesOn by default, free4 to 15 s
Seedance 2.0 Miniseedance-2.0-mini480p $0.047 · 720p $0.10Prompt, first and last frame, or up to 9 reference imagesOn by default, free4 to 15 s
Gemini Omni 1.2 Flashgemini-omni-1.2-flash720p $0.042 · 1080p $0.042 · 4K $0.07Prompt, first and last frame, or up to 6 reference imagesAlways on4 to 15 s
Kling 4.0kling-4.0720p $0.15 · 1080p $0.225Prompt, first and last frame, or up to 10 reference imagesAlways on3 to 30 s
Kling 3.0kling-3.0720p $0.185 (silent $0.125) · 1080p $0.235 (silent $0.18)Prompt or first and last frameOn by default; silent costs less3 to 15 s
Gemini Omni Flashgemini-omni-flash360p $0.042 · 720p $0.042 · 1080p $0.042 · 4K $0.07Prompt, first and last frame, or up to 10 reference imagesAlways on4 to 10 s
Wan 3.0wan-3.0480p $0.039 · 720p $0.079 · 1080p $0.158Prompt, first and last frame, or up to 10 reference imagesOn by default, free2 to 30 s
Wan 3.0 Primewan-3.0-prime480p $0.06 · 720p $0.123 · 1080p $0.246Prompt, first and last frame, or up to 10 reference imagesOn by default, free2 to 30 s
Veo 3.1 Fastveo-3.1-fast720p $0.20 per 8 s clipPrompt or a first frameAlways on8 s
Grok Imagine Video 1.5grok-imagine-video-1.5480p $0.071 · 720p $0.124Prompt, a first frame, or up to 3 reference imagesOn by default, free1 to 15 s
Muse Videomuse-video720p $0.02Prompt and up to 0 reference imagesAlways on5 to 10 s
Vidu Q3 Turbovidu-q3-turbo540p $0.03 · 720p $0.047 · 1080p $0.057Prompt or first and last frameOn by default, free1 to 16 s
HappyHorse 1.1happyhorse-1.1480p $0.059 · 720p $0.118 · 1080p $0.157Prompt, a first frame, or up to 9 reference imagesAlways on3 to 15 s
PixVerse V6pixverse-v6360p $0.027 (silent $0.019) · 540p $0.035 (silent $0.027) · 720p $0.047 (silent $0.035) · 1080p $0.09 (silent $0.07)Prompt or first and last frameOn by default; silent costs less1 to 15 s
MiniMax H3minimax-h3768p $0.039 · 2K $0.064Prompt and up to 9 reference images, 3 videos and 3 audio clipsAlways on4 to 15 s

Start, poll, download

import os, time, requests
API = "https://api.zurelay.com/v1"
HEADERS = {"Authorization": f"Bearer {os.environ['ZURELAY_API_KEY']}"}
# 1. Start the video
response = requests.post(f"{API}/videos", headers=HEADERS, json={
"model": "seedance-2.0",
"prompt": "A paper boat drifting down a rain-soaked street at night, neon reflections, slow dolly shot",
"seconds": 5,
"resolution": "720p",
"aspect_ratio": "16:9",
})
response.raise_for_status() # a 4xx says what to fix
job = response.json()
# 2. Poll until it's done (usually 2 to 6 minutes)
while job["status"] in ("queued", "in_progress"):
time.sleep(10)
job = requests.get(f"{API}/videos/{job['id']}", headers=HEADERS).json()
if job["status"] == "failed":
raise SystemExit(job["error"]["message"])
# 3. Download the MP4
video = requests.get(f"{API}/videos/{job['id']}/content", headers=HEADERS)
open("boat.mp4", "wb").write(video.content)

Every video model works the same way: change model and the request stays the same. The model guides cover what each one is best at: Seedance, Kling, Wan, Grok Imagine, Omni Flash, MiniMax H3, Veo, Vidu, PixVerse, Muse Video and HappyHorse.

Rather not poll? Add callback_url and the finished video is POSTed to your server, signed, the moment it’s ready. See Callbacks.

Parameters

modelstringrequired
A video model ID.
promptstring
What happens, up to 8,000 characters (Kling reads up to 2,500, Vidu reads up to 2,000, HappyHorse reads up to 5,000, PixVerse reads up to 2,048, MiniMax reads up to 7,000). Required unless you give a first frame or reference images.
negative_promptstring
What to keep out of the video, up to 2,500 characters. Kling and PixVerse read it; other models refuse it.
secondsinteger
Length in whole seconds, within the model’s range (see the table). Default 5, or the nearest length the model makes.
resolutionstring
Such as 480p, 720p, 1080p, 2k or 4k: each model’s are in the table. Default 720p, or the model’s lowest.
aspect_ratiostring
16:9, 9:16, 1:1, 4:3, 3:4, 21:9, or adaptive (the first frame’s shape; the default when you give one). Otherwise the default is 16:9. Omni, Veo and Muse make 16:9 and 9:16; Kling makes 16:9, 9:16, 1:1 and 21:9; Kling makes 16:9, 9:16 and 1:1; Grok, Vidu, HappyHorse and PixVerse make 16:9, 9:16, 1:1, 4:3 and 3:4.
sizestring
Instead of resolution and ratio: pixels like 1280x720.
generate_audioboolean
Sound and speech made with the video. Default true. (audio works too.) Kling and PixVerse cost less without sound.
seedinteger
The same seed and settings give similar results. Kling, Veo, Grok and Muse have no seed and ignore it.
first_framestring
An image the video starts on: a link or a data URL. MiniMax takes none: give the image in reference_images. Muse takes no images at all.
last_framestring
An image it ends on. Needs a first frame. Seedance, Omni, Kling, Wan, Vidu and PixVerse take one.
reference_imagesarray
Images of characters, products, places or a style to use: Seedance up to 9, Omni up to 10, Kling up to 10, Wan up to 10, Grok up to 3, HappyHorse up to 9, MiniMax up to 9; Veo, Muse, Vidu and PixVerse take none. Not with frames.
reference_videosarray
Videos whose look, motion or camera work to use, or a video to edit or extend: MP4 or MOV, as links or data URLs, mentioned in the prompt as @Video1, @Video2 and so on. Seedance 2.5, Seedance 2.0, Gemini Omni 1.2 Flash, Gemini Omni Flash and MiniMax H3 only. Their seconds are billed at part of a second’s price, or not at all on some models; see Reference videos and audio.
reference_audioarray
Voice, music or sound to use: MP3 or WAV, as links or data URLs, mentioned as @Audio1 and so on. Seedance 2.5, Seedance 2.0 and MiniMax H3 only, at no extra charge.
reference_modestring
With reference videos: edit one, extend one, or make a new video from the references (reference). Default auto: the model goes by the prompt. Seedance 2.5 only.
callback_urlstring
An https URL to send the finished (or failed) video to, instead of polling. See Callbacks.

Starting from images

First and last frame

Animate a still: the video opens exactly on first_frame. Add last_frame and it lands on that image, so you control both ends of the shot. MiniMax starts from reference images instead. Muse makes video from text only.

Request body
{
"model": "seedance-2.0",
"first_frame": "https://example.com/storefront-day.jpg",
"last_frame": "https://example.com/storefront-night.jpg",
"prompt": "Time passes from afternoon to night; lights come on inside the shop.",
"seconds": 6,
"resolution": "720p"
}

Reference images

Give up to 10 images (Seedance up to 9, Omni up to 10, Kling up to 10, Wan up to 10, Grok up to 3, HappyHorse up to 9, MiniMax up to 9; Veo, Muse, Vidu and PixVerse take none) and mention them in the prompt by order (@Image1, @Image2…): the people, products and places in them appear in the video.

Request body
{
"model": "seedance-2.5",
"reference_images": [
"https://example.com/chef.jpg",
"https://example.com/kitchen.jpg",
"https://example.com/dish.jpg"
],
"prompt": "The chef from @Image1 plates the dish from @Image3 in the kitchen from @Image2, close-up, warm light.",
"seconds": 8,
"resolution": "720p",
"aspect_ratio": "9:16"
}

Reference videos and audio

Give videos to borrow a person’s look and moves, a camera move or a scene, and audio for a voice, a tune or sound effects. Mention them in the prompt by order, as with images: @Video1, @Audio1. They mix with reference images, but not with a first frame.

Request body
{
"model": "seedance-2.5",
"reference_images": ["https://example.com/singer.jpg"],
"reference_videos": ["https://example.com/dance-moves.mp4"],
"reference_audio": ["https://example.com/chorus.mp3"],
"prompt": "The singer from @Image1 performs the dance from @Video1 on a rooftop at dusk, singing to @Audio1.",
"seconds": 8,
"resolution": "720p"
}
ModelReference videosReference audioEach second of reference video
Seedance 2.5Up to 10, 30 s in allUp to 10, 30 s in all480p $0.081, 720p $0.183
Seedance 2.0Up to 3, 15 s in allUp to 3, 15 s in all, with an image or video480p $0.0582, 720p $0.1194
Gemini Omni 1.2 FlashUp to 3, 15 s eachNoneNo charge
Gemini Omni FlashUp to 3, 10 s eachNoneNo charge
MiniMax H3Up to 3, 15 s eachUp to 3, 15 s in allNo charge

The model reads every second of a reference video, so those seconds are billed too, at part of a second’s price: a 5 s video with a 10 s reference costs its 5 seconds plus 10 reference seconds. Reference audio costs nothing more. Each video object shows the reference seconds billed (reference_seconds).

Editing and extending a video

Set reference_mode to edit to change a reference video (swap a person or object, add or remove something, change the music), or to extend to continue it forward or back. Say what to do in the prompt: “Replace the car in @Video1 with the one in @Image1”, “Continue @Video1 as the dancer leaves the stage”.

  • An edit keeps the video’s shape and length, so leave out aspect_ratio and seconds. It’s priced as long as the longest reference video, which must run 4 to 30 seconds.
  • An extension keeps the video’s shape; seconds sets how long the new video runs.
  • Without reference_mode the model decides from the prompt, and a prompt that reads as an edit with seconds or aspect_ratio set can fail (failed videos are free).

Rules for images

  • PNG, JPEG or WebP, as a public https link or a data URL, up to 10 MB.
  • 300 to 6000 pixels on each side, and no wider or taller than 5:2. Data URLs outside these are refused up front with a message saying why. Links are read when the video starts: one that isn’t a usable image fails the video with invalid_image, free of charge.
  • Frames or references, not both in one request.
  • Requests are JSON: images go in as links or data URLs, not as multipart uploads.

The video object

GET /v1/videos/video_abc123
{
"id": "video_abc123",
"object": "video",
"model": "seedance-2.0",
"status": "completed",
"progress": 100,
"created_at": 1790870400,
"completed_at": 1790870562,
"expires_at": 1793462400,
"seconds": 5,
"resolution": "720p",
"aspect_ratio": "16:9",
"size": "1280x720",
"generate_audio": true,
"url": "https://...signed link...",
"url_expires_at": 1790956962,
"cost": 1.035,
"error": null
}
FieldMeans
statusqueued, in_progress, completed or failed.
progressAn estimate from 0 to 100, from how long the model usually takes.
urlThe MP4, once completed. The link works for 24 hours; fetch the job again for a fresh one.
expires_atWhen the video is deleted: 30 days after it was made, or sooner if someone deletes it in the Library.
costWhat it cost, once completed.
reference_secondsSeconds of reference video in the price, when you gave some.
errorFor failed videos: code (content_policy, invalid_image, invalid_request when the model can’t take the request as asked, generation_failed, or generation_timeout after 90 minutes) and a message.

Other endpoints

EndpointReturns
GET /v1/videos/{id}One video, as above.
GET /v1/videos/{id}/contentThe MP4 itself.
GET /v1/videos?limit=20&after=video_abcYour workspace’s videos, newest first, with first_id and last_id for paging.

Paying for video

A video costs its price per second (by resolution) times its length; Kling and PixVerse cost less per second without sound; reference videos add their seconds at part of that price. When you start one, its price is set aside from your balance; it’s charged when the video is done, and released if it fails. A failed video is free.

Draft at the lowest resolution and a short length, then make the final cut at a higher resolution once the prompt is right. Most of the wait is the queue, so drafts aren’t much faster, but they cost a fraction.

Questions, or something missing? Ask our support team and we’ll answer by email, usually within a minute.