What is a video generation API?

A video generation API creates a video clip from a text prompt, and on many models from images, video or audio you supply. You submit a job, poll until the video is ready, and most models bill per second of output at a rate set by the resolution.

Model types

What a video generation API does

A video generation API turns a description into a short video clip. The simplest mode is text-to-video: you describe the scene, the action and the camera movement, and the model renders the clip. Most video models also accept other inputs:

  • Image-to-video. An image becomes the first frame, and the model animates it. Some models also take a last frame and generate the motion between the two. See image-to-video.
  • Reference-to-video. Reference images, and on some models video or audio clips, carry characters, products or a style into a new scene instead of becoming its first frame.
  • Editing and extension. Some models edit or restyle an existing clip, or continue it.

Many models also generate a synchronized audio track, such as dialogue, sound effects or music, and some let you turn it off. Teams use video generation for ads and social clips, product demos, explainers, storyboards and concept previews, and to animate existing artwork.

How a request works: an asynchronous job

Video generation runs as an asynchronous job. You send POST /v1/videos/generations with the model ID, the prompt and settings such as duration, resolution and aspect_ratio. The API answers at once with a job_id and a polling URL. You then call GET /v1/jobs/{job_id} until the status is completed, and read the video URL from the result. A job that does not succeed ends as failed with an error.

Jobs typically finish in 1 to 15 minutes, and 4K output can take 30 minutes or more. To continue a clip you generated, send extend_from with its job ID or a video URL.

Settings that shape the clip

  • Duration. Set in seconds. Most models generate clips of a few seconds up to 15 seconds, and some reach 30 seconds or more.
  • Resolution. Usually between 480p and 1080p. Some models also offer 360p, and some go up to 2K or 4K.
  • Aspect ratio. Text-to-video takes a ratio such as 16:9, 9:16 or 1:1. Image-to-video usually follows the shape of the source image.
  • Frame rate. Set by the model, not by a parameter. Several models render at 24 fps.
  • Audio. On models with native audio, a setting such as generate_audio turns the soundtrack on or off.
  • Seed and negative prompt. Some models accept a seed for repeatable results and a negative prompt that lists what to avoid.

Choosing a model and understanding billing

Run the same prompt through a few candidates and compare motion quality, prompt adherence and how well characters stay consistent. Then check the limits that matter for your project: maximum duration and resolution, supported inputs, native audio and generation time. Side-by-side model comparisons show two models' capabilities and rates together.

Most video models bill per second of generated video, at a rate set by the output resolution, so cost grows with both length and resolution: a 10-second clip costs twice as much as a 5-second clip at the same settings. Some models charge a different rate with audio on or off, a few bill a fixed amount per video for a set length and resolution, and requests that edit or reference a source video can bill the input seconds as well as the output.

Try and call a video model on EmpirioLabs AI

Browse the video models and run a prompt in the Playground before you write code. To call a model, create an API key in the dashboard, submit the job to https://api.empiriolabs.ai/v1/videos/generations and poll the jobs endpoint. The video generation reference lists every request field, and generation templates apply curated effects with a recommended model and settings.

Pay as you go, or choose a plan with weekly allowances. Current rates for every model are on the pricing page.

Video Generation models on EmpirioLabs AI

37 models, cheapest first where they share a billing unit. Each model's page has its full rate card, parameters and code.

Endpoints

POST /v1/videos/generations

Priced per second

ModelLowest rateApplies to
Wan 2.2 Alibaba Cloud$0.015 per secondFlash 480P
Wan 2.6 Alibaba Cloud$0.0250$0.0225 per secondFlash 720P (no audio)
PixVerse V6 Max PixVerse$0.030667$0.024534 per secondStandard 360p no audio
MiniMax H3 Lite MiniMax$0.025 per second480p
Pixverse Avatar PixVerse$0.025556 per second360p
Pixverse v6 PixVerse$0.025556 per secondStandard 360p no audio
Pixverse C1 PixVerse$0.030667 per secondStandard 360p no audio
Wan 2.1 Alibaba Cloud$0.036 per secondTurbo
Show all 30 models
ModelLowest rateApplies to
Wan 2.5 Alibaba Cloud$0.05 per second480P
SVI 2.0 Pro VITA-Group / EPFL$0.057 per second480p Video
MiniMax H3 Highspeed MiniMax$0.06 per second768p
Hunyuan Video 1.5 Tencent$0.075$0.061 per second480p
Seedance 2.0 Mini ByteDance$0.070 per secondT2V/I2V 480P
Wan 3.0 Alibaba Cloud$0.07 per secondStandard 480P
Kling 4.0 Flash Kling AI$0.094 per secondStandard 720p
Grok Imagine Video 1.5 xAI$0.096 per second480p
Wan 2.7 Alibaba Cloud$0.10 per secondAll Modes 720P
MOSS Video and Audio OpenMOSS$0.12 per second360p Video
Seedance 2.0 Fast ByteDance$0.122 per secondT2V/I2V 480P
Seedance 2.5 Flash ByteDance$0.135 per secondT2V/I2V 480P
Seedance 2.0 Pro ByteDance$0.139 per secondT2V/I2V 480P
HappyHorse 1.0 Alibaba Cloud$0.14 per secondAll Modes 720P
HappyHorse 1.1 Alibaba Cloud$0.14 per second720p
Kling v3 Motion Control Kling AI$0.14 per secondStandard (720p)
Kling 4.0 Kling AI$0.157 per secondStandard 720p
Kling O3 Kling AI$0.168 per secondStandard T2V/I2V
Kling V3 Kling AI$0.168 per secondStandard T2V/I2V
Kling 3.0 Turbo Kling AI$0.18 per second720p
MiniMax H3 MiniMax$0.18 per second768p
Seedance 2.5 ByteDance$0.206 per secondT2V/I2V 480P

Priced per video

ModelLowest rateApplies to
Pixverse v5 PixVerse$0.23 per video360p/540p 5s
Pixverse v5.6 PixVerse$0.35 per video360p/540p 5s no audio

Also available: Pixverse Lip Sync, Pixverse Edit, Sonilo Video to Video, Sonilo Video Translation, Sonilo Video Analysis

Frequently asked questions

How long does it take to generate a video with an API?

Video jobs typically finish in 1 to 15 minutes, depending on the model and settings, and 4K output can take 30 minutes or more. The API returns a job ID at once, and you poll it until the video is ready.

How is AI video generation priced?

Most models bill per second of generated video, with the rate set by resolution, so longer and sharper clips cost more. Some models charge a different rate with audio on or off. On EmpirioLabs AI you can pay as you go or choose a plan with weekly allowances.

What is the difference between text-to-video and image-to-video?

Text-to-video creates the whole clip from a prompt. Image-to-video starts from an image you supply, usually as the first frame, so you control the subject and composition before any motion is added.

Can AI video models generate sound?

Many can. Models with native audio generate a synchronized soundtrack, such as dialogue, sound effects or music, and some let you turn it off.

What resolutions can AI video models produce?

Depending on the model, from 360p up to 1080p, and 2K or 4K on some. Higher resolutions cost more per second and can take longer to render.

Which Video Generation model costs the least on EmpirioLabs AI?

Wan 2.2 lists the lowest rate per second of the 30 Video Generation models priced that way, followed by Wan 2.6, PixVerse V6 Max. The price table on this page shows each model's lowest rate and what it applies to.

Ready to use better endpoints?

Check out our pricing or reach out if you want your own model deployed on our stack.