Skip to main content
POST

Choose a model

The API supports MiniMax H3 Max, MiniMax H3, Seedance 2.5, Gemini Omni Flash 1.1, FLUX 3, WAN 3.0, Grok Imagine 1.5, and other video families. Use the video model reference to choose an exact variant and its supported inputs, durations, and resolutions.
The REST API requires model. The MCP generate_video tool defaults to MiniMax H3 Max when you omit it. REST fields use camelCase, such as aspectRatio and referenceVideos; MCP uses aspect_ratio and reference_videos.

Create a video

Send one generation request and save its returned id. Generation runs asynchronously; an accepted request does not mean the video is ready. Most videos take a few minutes, and queues can take longer.
MiniMax H3 Max accepts 5–15 whole seconds and 480p or 768p. MiniMax H3 accepts 4–15 whole seconds and 768p or 2k. Both families generate native audio; generateAudio: false does not disable it. Models with an audio toggle, such as Seedance 2.5, can accept generateAudio: false. Providing reference audio enables audio generation.
Send duration as a string. Veo 3.1 text variants accept "4s", "6s", or "8s"; their Image and First-Last Frame variants accept only "8s". Veo 3.1 does not accept five seconds. Equivalent strings such as "8" and "8s" are normalized by the API.

Animate an image

Select a model with a first-image input and send image. Adding an image does not automatically switch a text model to its image variant.
Add "tailImage": "https://yourcdn.com/last-frame.png" when you want a last frame and the chosen variant supports it. Check the Input column in the model catalog; model suffixes alone do not identify every input mode.

Use reference media

Reference variants use images to preserve subjects or style. Some also accept videos for motion or editing and audio for voice or sound guidance. This example uses a character image, a motion clip, and a voice clip. Replace the placeholder URLs with your own media. Keep each MiniMax H3 Max reference video and audio clip between 2 and 15 seconds, and each type’s combined duration at or below 15 seconds.
  • Use referenceImages, referenceVideos, and referenceAudios as arrays of { "url": "..." } objects.
  • referenceVideo is a single-URL alternative. If you supply the referenceVideos array, it takes precedence, including when empty.
  • Do not combine reference videos with image or tailImage. Audio references require a reference image or a supported reference video; they cannot accompany a first-frame image.
  • Limits differ by variant. MiniMax H3 Max Reference supports up to 9 images; Seedance 2.5 Reference supports up to 30. See reference limits for counts, durations, formats, and sizes.
  • referenceImages[].type is optional. Kling O1/O3 uses character, object, or style to assign reference roles; other families use general image references.
All input URLs must be downloadable server-side without browser cookies. A URL that works in your browser may still reject the provider with HTTP 403. Signed URLs must remain valid through queue delays and the provider’s download. Reference media can increase credit usage.

Poll for the result

Call Get Video with the returned id and the same account’s API key every 15–30 seconds. Read video.url when video.status is completed. The following Node.js 20+ example checks HTTP errors, handles malformed responses, and stops after ten minutes. A local timeout does not cancel the generation; keep the ID so you can check it later.

Handle errors and access limits

Get Video currently exposes only processing and completed, so do not wait for a failed status. Submitting a new generation can create another billable job; a polling timeout is not a reason to recreate it automatically.

Authorizations

x-api-key
string
header
required

API key for authentication. Get yours at https://easy-peasy.ai/settings/api

Body

application/json
prompt
string
required

Text prompt describing the video to generate.

Minimum string length: 1
Example:

"A cat walking on the beach at sunset, cinematic lighting"

model
enum<string>
required

Required exact model title. See the video model reference for input modes, durations, resolutions, and reference limits. Adding media does not automatically switch model variants. https://docs.easy-peasy.ai/api-reference/video-models

Available options:
Veo 3.1 Fast,
Gemini Omni Flash,
Gemini Omni Flash Reference,
Gemini Omni Flash 1.1,
Gemini Omni Flash 1.1 Reference,
MiniMax H3 Max,
MiniMax H3 Max Reference,
Grok Imagine,
Grok Imagine Reference,
Grok Imagine 1.5 Text,
Grok Imagine 1.5 Reference,
Seedance 2.0 Fast,
MiniMax H3,
Seedance 2.5,
Seedance 2.5 Turbo,
Seedance 2.5 Turbo Reference,
Seedance 2.5 Reference,
FLUX 3,
WAN 3.0,
WAN 3.0 Reference,
Seedance 2.0 Turbo,
Seedance 2.0,
Seedance 2.0 Mini,
Seedance 2.0 Reference,
Seedance 2.0 Fast Reference,
MiniMax H3 Reference,
Seedance 2.0 Mini Reference,
Kling 2.6 Pro,
Veo 3.1 Lite,
Veo 3.1 Lite Image,
Veo 3.1 Fast Image,
Gemini Omni Flash Image,
Gemini Omni Flash 1.1 Image,
Gemini Omni Flash 1.1 First-Last Frame,
MiniMax H3 Max Image,
Grok Imagine Image,
Grok Imagine 1.5,
Seedance 2.0 Image,
Seedance 2.0 Fast Image,
MiniMax H3 Image,
Seedance 2.5 Turbo Image,
Seedance 2.5 Image,
FLUX 3 Image,
WAN 3.0 Image,
Seedance 2.0 Turbo Image,
Seedance 2.0 Mini Image,
Happy Horse,
Happy Horse Image,
Happy Horse Reference,
Kling 2.6 Pro Image,
Kling 3.0 Pro,
Kling 3.0 Pro Image,
Kling 3.0 Standard,
Kling 3.0 Standard Image,
Kling O3 Pro,
Kling O3 Pro Image,
Kling O3 Standard,
Kling O3 Standard Image,
Kling O3 Reference,
Seedance 1.5 Pro,
Seedance 1.5 Pro Image,
Kling O1 Image,
Kling O1 Reference,
Kling 2.5 Turbo Pro Image,
Kling 2.5 Turbo Pro,
Kling 2.5 Turbo Standard,
Seedance 1.5 Pro First-Last Frame,
Veo 3.1,
LTX-2 Pro,
LTX-2 Fast,
Veo 3.1 Image,
Veo 3.1 First-Last Frame,
Veo 3.1 Fast First-Last Frame,
LTX-2 Pro Image,
LTX-2 Fast Image,
Hailuo 2.0,
Hailuo 2.0 Pro,
Hailuo 2.3,
Hailuo 2.3 Pro,
Hailuo 2.3 Image,
Hailuo 2.3 Pro Image,
Hailuo 2.3-Fast Pro Image,
Seedance v1 Pro Fast,
Pixverse v5.5,
Wan 2.5,
Wan v2.2 Turbo,
Wan v2.2,
Hailuo 2.0 Image,
Hailuo 2.0 Pro Image,
Seedance 1.0 Pro Fast,
Pixverse v5.5 Image,
Wan 2.5 Image,
Wan v2.2 Turbo Image,
Wan-2.2 Image
Example:

"MiniMax H3 Max"

image
string<uri>

Public URL of a first-frame image. Choose a variant with a first-image input in the video model reference. Model suffixes are not always enough: Grok Imagine 1.5 and Kling 2.5 Turbo Standard also require image. The URL must remain downloadable without browser cookies through queue delays.

duration
string

Output length as a string, using the exact variant’s allowed durations. MiniMax H3 Max: 5–15 whole seconds; MiniMax H3: 4–15. Veo 3.1 text variants: 4s, 6s, 8s; Image and First-Last Frame variants: 8s only. Bare seconds and an s suffix are equivalent. If omitted, the API selects five seconds when listed, otherwise the first configured duration, or five seconds for models without a list. Provider-specific output lengths can differ.

Example:

"5"

aspectRatio
enum<string>
default:16:9

Use an aspect ratio listed for the exact variant. Omit for variants whose framing follows the input. The API defaults to 16:9 when applicable.

Available options:
16:9,
9:16,
1:1,
4:3,
3:4,
21:9,
3:2,
2:3
resolution
enum<string>

Allowed resolutions depend on the exact variant. MiniMax H3 Max: 480p/768p; MiniMax H3: 768p/2k; Seedance 2.5: 480p/720p; Seedance 2.5 Turbo: 720p/1080p. Send a listed value explicitly; defaults vary by model. Use lowercase k for 2k and 4k. Omit when no resolution selector is listed.

Available options:
360p,
480p,
720p,
768p,
1080p,
2k,
4k
generateAudio
boolean

Enable or disable generated audio only on models with an audio toggle. MiniMax H3 and H3 Max always generate native audio; false does not disable it. Supplying reference audio enables audio generation. Audio settings can affect credit usage.

tailImage
string<uri>

Public URL of a last-frame image, paired with image. Use only a variant that supports a last image, such as MiniMax H3 Max Image or Seedance 2.5 Image. See the video model reference.

referenceImages
object[]

Array of reference images on supported variants. Reference models require at least one reference image or a supported reference video. Limits differ: MiniMax H3 Max Reference accepts up to 9 images; Seedance 2.5 Reference accepts up to 30. Some Veo variants also support optional references. See the video model reference.

referenceVideo
string<uri>

Single reference video URL, on variants supporting reference video. Alternative to referenceVideos; the array takes precedence when supplied, even if empty.

referenceVideos
object[]

Reference videos for supported variants only. Do not combine with image or tailImage. Counts, clip lengths, total duration, and file-size limits depend on the variant. MiniMax H3 Max Reference accepts up to 3 clips of 2–15 seconds, at most 15 seconds combined, and 20 MiB each. Seedance 2.5 Reference supports up to 3 clips of 2–30 seconds, at most 30 seconds combined, and 50 MiB each. Reference inputs can add credit charges.

referenceAudios
object[]

Audio references for supported variants. Require a companion reference image or supported reference video, and cannot accompany image. Enable generated audio. MiniMax H3 Max Reference accepts up to 3 clips of 2–15 seconds and at most 15 seconds total. Seedance 2.5 Reference supports up to 10 MP3/WAV clips, 2–30 seconds each, at most 30 seconds total, and 15 MiB each. See the reference limits for other variants.

Response

Video generation started

id
integer

Video ID. Use this to poll for the result with the Get Video endpoint.

prompt
string

The prompt used for generation

image_url
string

Video URL when present. May be empty or omitted while processing. Poll Get Video with id; do not depend on this field to detect submission success.

model
string

The model used for generation

is_video
boolean
created_at
string<date-time>