Video Generation

Overview

LemonData provides access to 69+ video generation models from 11 providers through a single unified API. Video generation is asynchronous — you submit a request and receive a task ID, then poll for the result.

The model list is updated frequently. For the latest available models and pricing, visit the Models page or use the Models API.

Async Workflow

import requests
import time

API_KEY = "sk-your-api-key"
BASE = "https://api.lemondata.cc/v1"
headers = {"Authorization": f"Bearer {API_KEY}"}

# Step 1: Submit generation request
resp = requests.post(f"{BASE}/videos/generations",
    headers=headers,
    json={
        "model": "kling-v2.6-pro",
        "prompt": "A golden retriever running on a beach at sunset, cinematic 4K",
        "duration": 5,
        "aspect_ratio": "16:9"
    }
)
task_id = resp.json()["task_id"]

# Step 2: Poll for result
while True:
    status = requests.get(f"{BASE}/videos/generations/{task_id}", headers=headers).json()
    if status["status"] in ("completed", "succeeded"):
        print(f"Video URL: {status['video_url']}")
        break
    elif status["status"] == "failed":
        print(f"Failed: {status.get('error')}")
        break
    time.sleep(10)

Model Capability Matrix

Different models excel at different tasks. Use this matrix to choose the right model for your use case. Legend: ✅ Supported | ❌ Not supported

Series	Provider	T2V	I2V	Keyframe	Extension	Editing	Max Duration
Sora	OpenAI	✅	❌	❌	❌	❌	~20s
Kling	Kuaishou	✅	✅	❌	✅	❌	10s
Veo	Google	✅	❌	❌	❌	❌	8s
Seedance	ByteDance	✅	✅	✅	✅	✅	10s
Hailuo	MiniMax	✅	✅	❌	❌	❌	6s
Wan	Alibaba	✅	✅	❌	❌	❌	5s
Runway	Runway	✅	✅	❌	❌	❌	10s
Luma	Luma	✅	❌	❌	✅	❌	5s
Vidu	Vidu	✅	✅	❌	❌	❌	8s
Grok	xAI	✅	❌	❌	❌	❌	~10s
Higgsfield	Higgsfield	✅	❌	❌	❌	❌	~5s

Capability Definitions

T2V (Text-to-Video): Generate video from a text prompt
I2V (Image-to-Video): Animate a static image into video using image_url or image
Keyframe: Control start and end frames with start_image + end_image
Extension: Extend an existing video’s duration
Editing: Modify specific aspects of an existing video

Available Models by Series

Sora (OpenAI)

Model	Quality	Notes
`sora-2`	Standard	Default model, good balance of quality and speed
`sora-2-pro`	High	Higher quality, longer generation time
`sora-2-characters`	Standard	Character-focused generation

Kling (Kuaishou)

Model	Capability	Notes
`kling-v2.6-pro`	T2V	Latest generation, professional quality
`kling-v2.6-std`	T2V	Latest generation, fast
`kling-v2.5-turbo-pro`	T2V	Turbo speed, pro quality
`kling-v2.1-master`	T2V/I2V	Master quality
`kling-v2.1-pro`	T2V/I2V	Professional quality
`kling-v2.1-standard`	T2V/I2V	Standard quality
`kling-video`	T2V/I2V	Base model
`kling-video-extend`	Extension	Extend existing videos
`kling-video-o1-pro`	T2V	O1 reasoning, pro quality
`kling-video-o1-std`	T2V	O1 reasoning, standard
`kling-effects`	Effects	Apply visual effects
`kling-omni-video`	T2V	Omni model
`kling-motion-control`	T2V	Motion-controlled generation

Veo (Google)

Model	Quality	Notes
`veo3.1`	Standard	Google’s latest video model
`veo3.1-pro`	High	Professional quality
`veo3.1-4k`	Ultra	4K resolution output
`veo3.1-fast`	Fast	Faster generation
`veo3.1-fast-4k`	Fast + 4K	Fast generation with 4K output
`veo3.1-components`	Standard	Component-based generation
`veo3`	Standard	Previous generation
`veo3-pro`	High	Previous gen, professional
`veo3-fast`	Fast	Previous gen, fast

Seedance (ByteDance)

Model	Capability	Notes
`seedance-2-0`	T2V/I2V/Keyframe/Extension/Editing	Latest, most capable
`seedance-1-5-pro`	T2V/I2V	Previous generation, pro quality
`seedance-1-0-pro`	T2V/I2V	First generation, pro
`seedance-1-0-pro-fast`	T2V/I2V	First generation, fast
`seedance-1-0-lite-t2v`	T2V	Lightweight text-to-video
`seedance-1-0-lite-i2v`	I2V	Lightweight image-to-video

Seedance 2.0 supports the widest range of capabilities including multimodal-to-video, video extension, and video editing — all through the same API endpoint.

Hailuo (MiniMax)

Model	Quality	Notes
`hailuo-2.3`	Standard	Good quality
`hailuo-2.3-pro`	High	Higher quality output
`hailuo-2.3-fast`	Fast	Faster generation
`hailuo-2.3-standard`	Standard	Standard tier
`video-01`	Standard	MiniMax video-01
`video-01-live`	Standard	Live-style generation

Wan (Alibaba)

Model	Capability	Notes
`wan-2.6`	T2V	Latest text-to-video
`wan2.6-i2v`	I2V	Latest image-to-video
`wan-2.5`	T2V	Previous generation
`wan2.5-i2v-preview`	I2V	Previous gen I2V
`wan-2.2-plus`	T2V	Earlier generation
`vace-14b`	T2V	VACE architecture

Runway

Model	Duration	Notes
`runwayml-gen4-turbo-5`	5s	Fast generation
`runwayml-gen4-turbo-10`	10s	Longer clips

Luma

Model	Capability	Notes
`luma-video-api`	T2V	Text-to-video
`luma-video-extend-api`	Extension	Extend existing videos

Vidu (Shengshu)

Model	Quality	Notes
`viduq3-pro`	High	Latest generation
`viduq2-pro`	High	Previous gen, pro
`viduq2-pro-fast`	Fast	Previous gen, fast pro
`viduq2`	Standard	Previous gen, standard
`viduq2-turbo`	Fast	Turbo speed
`vidu2.0`	Standard	Base model

Grok (xAI)

Model	Notes
`grok-video-3`	xAI’s video generation model
`grok-video-3-10s`	10-second variant

Higgsfield

Model	Notes
`higgsfield-turbo`	Fastest, lower cost
`higgsfield-standard`	Standard quality
`higgsfield-lite`	Lightweight

Usage Examples

Text-to-Video (T2V)

The most common use case. All models support this.

response = requests.post(f"{BASE}/videos/generations",
    headers=headers,
    json={
        "model": "veo3.1-pro",
        "prompt": "Aerial drone shot of a coastal city at golden hour, waves crashing against cliffs",
        "duration": 5,
        "aspect_ratio": "16:9",
        "resolution": "1080p"
    }
)

Image-to-Video (I2V)

Animate a static image. Use image_url for a URL or image for base64 data.

# Using image URL
response = requests.post(f"{BASE}/videos/generations",
    headers=headers,
    json={
        "model": "wan2.6-i2v",
        "prompt": "The person slowly turns and smiles at the camera",
        "image_url": "https://example.com/portrait.jpg"
    }
)

# Using base64 image
import base64
with open("photo.jpg", "rb") as f:
    image_b64 = base64.b64encode(f.read()).decode()

response = requests.post(f"{BASE}/videos/generations",
    headers=headers,
    json={
        "model": "kling-v2.1-master",
        "prompt": "Gentle wind blows through the scene",
        "image": f"data:image/jpeg;base64,{image_b64}"
    }
)

Keyframe Control (Start + End Image)

Control both the first and last frames for precise transitions. Currently supported by Seedance 2.0.

response = requests.post(f"{BASE}/videos/generations",
    headers=headers,
    json={
        "model": "seedance-2-0",
        "prompt": "Smooth transition from day to night, city lights gradually turning on",
        "start_image": "https://example.com/city-day.jpg",
        "end_image": "https://example.com/city-night.jpg",
        "duration": 5
    }
)

Video Extension

Extend an existing video’s duration. Use models with extension capability.

response = requests.post(f"{BASE}/videos/generations",
    headers=headers,
    json={
        "model": "kling-video-extend",
        "prompt": "Continue the scene naturally",
        "image_url": "https://example.com/last-frame.jpg"
    }
)

Parameters Reference

Parameter	Type	Description
`model`	string	Model ID (default: `sora-2`)
`prompt`	string	Required. Text description of the video
`image_url`	string	URL of starting image (for I2V)
`image`	string	Base64-encoded image with data URL prefix (for I2V)
`duration`	integer	Video duration in seconds (1-60, model-dependent)
`aspect_ratio`	string	`16:9`, `9:16`, `1:1`, etc.
`resolution`	string	`1080p`, `720p`, `4k`
`fps`	integer	Frames per second (1-120)
`negative_prompt`	string	What to avoid in generation
`seed`	integer	Random seed for reproducibility
`cfg_scale`	number	Guidance scale (0-20)
`motion_strength`	number	Motion intensity (0-1)
`start_image`	string	URL of starting keyframe
`end_image`	string	URL of ending keyframe

Not all parameters are supported by every model. Unsupported parameters are silently ignored. Check the model’s documentation for supported parameters.

Model Selection Guide

Best Quality

Seedance 2.0 or Kling v2.6 Pro — cinematic quality, rich detail, natural motion

Fastest Generation

Higgsfield Turbo or Hailuo 2.3 — quick results for prototyping and iteration

Most Versatile

Seedance 2.0 — supports T2V, I2V, keyframe, extension, and editing in one model

Best Value

Wan 2.6 or Hailuo 2.3 — competitive quality at lower cost per generation

Billing

Video generation uses fixed per-generation pricing. You are charged once when the task is submitted, regardless of video duration. If generation fails, the charge is automatically refunded. Check current pricing on the Models page or via the Pricing API.

Getting Started

Guides

Overview

Async Workflow

Model Capability Matrix

Capability Definitions

Available Models by Series

Sora (OpenAI)

Kling (Kuaishou)

Veo (Google)

Seedance (ByteDance)

Hailuo (MiniMax)

Wan (Alibaba)

Runway

Luma

Vidu (Shengshu)

Grok (xAI)

Higgsfield

Usage Examples

Text-to-Video (T2V)

Image-to-Video (I2V)

Keyframe Control (Start + End Image)

Video Extension

Parameters Reference

Model Selection Guide

Best Quality

Fastest Generation

Most Versatile

Best Value

Billing

Getting Started

Guides

​Overview

​Async Workflow

​Model Capability Matrix

​Capability Definitions

​Available Models by Series

​Sora (OpenAI)

​Kling (Kuaishou)

​Veo (Google)

​Seedance (ByteDance)

​Hailuo (MiniMax)

​Wan (Alibaba)

​Runway

​Luma

​Vidu (Shengshu)

​Grok (xAI)

​Higgsfield

​Usage Examples

​Text-to-Video (T2V)

​Image-to-Video (I2V)

​Keyframe Control (Start + End Image)

​Video Extension

​Parameters Reference

​Model Selection Guide

Best Quality

Fastest Generation

Most Versatile

Best Value

​Billing

Overview

Async Workflow

Model Capability Matrix

Capability Definitions

Available Models by Series

Sora (OpenAI)

Kling (Kuaishou)

Veo (Google)

Seedance (ByteDance)

Hailuo (MiniMax)

Wan (Alibaba)

Runway

Luma

Vidu (Shengshu)

Grok (xAI)

Higgsfield

Usage Examples

Text-to-Video (T2V)

Image-to-Video (I2V)

Keyframe Control (Start + End Image)

Video Extension

Parameters Reference

Model Selection Guide

Billing