Skip to main content

​
Overview

The Text-to-Video API provides access to 40+ cutting-edge models that generate videos from text descriptions. From cinematic sequences to animated content, these models cover a wide range of video generation use cases.
Text-to-video models are automatically selected when no start frame image is provided in the Video Studio.

​
Model Categories

ByteDance’s video generation models with quality and resolution controls.
  • Seedance Lite — Fast generation for quick prototyping
  • Seedance Pro / Pro Fast — High-quality cinematic generation
  • Seedance v1.5 Pro / Pro Fast — Enhanced motion and detail
  • Seedance 2.0 — Latest generation with 15s duration support
  • Seedance 2.0 Extend — Seamlessly continue existing videos
Kuaishou’s photorealistic video generation models.
  • Kling v2.1 Master — High-fidelity cinematic rendering
  • Kling v2.5 Turbo Pro — Fast generation with quality
  • Kling v2.6 Pro — 5s/10s duration with ultra-realism
  • Kling O1 Pro — Latest generation with improved coherence
  • Kling v3.0 Pro / Standard — Next-gen detail and motion
Industry-leading commercial video generation models.
  • Veo 3 / 3 Fast — Google’s video generation model
  • Veo 3.1 / 3.1 Fast — Enhanced 8s generation at 1080p
  • Sora / Sora 2 / Sora 2 Pro — OpenAI’s flagship models (10–25s)
  • Runway Gen-3 — Slow-motion and cinematic effects
  • Hailuo 02 Standard / Pro — Minimax 6s/10s generation
  • Hailuo 2.3 Standard / Pro — Latest minimax models
Realistic character animation and motion models.
  • Wan 2.1 / 2.2 / 2.2 Fast — Portrait and character animation
  • Wan 2.5 / 2.5 Fast — Enhanced motion quality
  • Wan 2.6 — 5s/10s/15s duration support
Domain-specific and open-source video models.
  • Hunyuan / Hunyuan Fast — Tencent’s multilingual model
  • Pixverse v4.5 / v5 / v5.5 — Anime and illustration styles
  • Vidu v2.0 — 4s 9:16 mobile-optimized videos
  • OVI — Creative and experimental generation
  • Grok Imagine — X.AI’s video model with 6s/10s options
  • LTX 2 Pro / Fast / 19B — Open-source with long duration (20s)
ModelDurationResolutionAspect RatiosBest For
Seedance 2.05s / 10s / 15sBasic/High16:9, 9:16, 4:3, 3:4Production-ready cinematic videos
Kling v3.0 Pro5sVariable16:9, 9:16, 1:1Photorealistic motion
Veo 3.18s1080p16:9, 9:16Google-grade video quality
Sora 2 Pro10s / 15s / 25s720p/1080p16:9, 9:16Long-form storytelling
Runway Gen-35s / 8s720p/1080p5 ratiosCinematic slow-motion
Wan 2.65s / 10s / 15s720p/1080p16:9, 9:16Character animation
LTX 2 Fast6–20s720p16:9, 9:16Open-source long videos

​
Newly Added Models

These models were recently added to the platform and offer cutting-edge video generation capabilities.

​
Seedance 2.0

Family: ByteDance
Duration: 5s / 10s / 15s
Quality: Basic / High
Aspect Ratios: 16:9, 9:16, 4:3, 3:4
  • ByteDance’s latest video generation model
  • Up to 15 seconds of high-quality video
  • Quality presets for speed vs. fidelity tradeoff
  • Wide aspect ratio support
  • Natural motion and physics simulation
Example Request
{
  "model": "seedance-v2.0-t2v",
  "prompt": "A drone shot flying through a neon-lit cyberpunk city at night, with glowing signs reflecting on wet streets, moving traffic, and rain falling gently.",
  "aspect_ratio": "16:9",
  "duration": 10,
  "quality": "high"
}

​
Seedance 2.0 Extend

Family: ByteDance
Duration: 5s / 10s / 15s
Quality: Basic / High
Requires: request_id from original Seedance 2.0 generation
  • Seamlessly continue any Seedance 2.0 video
  • Preserves style, motion, and audio continuity
  • Optional continuation prompt for guided extension
  • Same quality and duration options as base model
Example Request
{
  "model": "seedance-v2.0-extend",
  "request_id": "abcdefg-123-456-789-a1b2c3d4e5f6",
  "prompt": "The camera continues to pull back, revealing the full cityscape under a starry sky.",
  "duration": 10,
  "quality": "high"
}
Seedance 2.0 Extend requires the request_id from a previously generated Seedance 2.0 video. The extension will match the style and motion of the original.

​
Common Input Parameters

​
Prompt

Type: string
Required: Yes
Max Length: Varies by model (typically 500–3000 characters) Detailed text description of the video. Include camera movements, subject actions, lighting, atmosphere, and style.
For best results, structure your prompt with:
  1. Subject/Scene: What is in the frame
  2. Action/Motion: What is moving or changing
  3. Camera: Camera movement and angle
  4. Lighting/Atmosphere: Mood and visual style
  5. Style: Cinematic, realistic, animated, etc.
Example: 
A lone astronaut walks slowly across a desolate Martian landscape at sunset. 
Red dust swirls around her boots as she approaches a distant research station. 
The camera follows behind her with a slow dolly, golden sunlight creating long shadows. 
Cinematic wide-angle shot, ultra-realistic, 8K quality, dramatic atmosphere.

​
Aspect Ratio

Type: string
Options: Varies by model (typically 16:9, 9:16, 1:1, 4:3, 3:4, 21:9) Defines the width-to-height ratio of the generated video. Most models support 16:9 (landscape) and 9:16 (vertical).

​
Duration

Type: integer
Options: Varies by model (typically 4–25 seconds) Length of the generated video in seconds. Longer durations consume more credits and take longer to generate.
Some models have fixed durations (e.g., Vidu v2.0 only supports 4s), while others offer flexible options (e.g., Wan 2.6 supports 5s/10s/15s).

​
Resolution

Type: string
Options: 480p, 720p, 1080p, 768P, etc. Output video resolution. Higher resolutions provide better quality but consume more credits.

​
Quality

Type: string
Options: basic, medium, high Some models (like Seedance and Wan) offer quality presets that affect generation time and fidelity.

​
Model-Specific Features

Durations: 10s, 15s, 25s
Resolutions: 720p, 1080p
Aspect Ratios: 16:9, 9:16
25-second videos only support 720p resolution.
Best for narrative storytelling and complex scene transitions.
Durations: 5s, 8s
Resolutions: 720p, 1080p
Aspect Ratios: 16:9, 9:16, 1:1, 4:3, 3:4
8-second videos cannot use 1080p resolution. 1080p videos cannot be 8 seconds.
Excels at slow-motion, cinematic camera work, and dramatic lighting.
  • LTX 2 Fast: 6–20 seconds (2s increments)
  • LTX 2 Pro: 6, 8, 10 seconds
  • LTX 2 19B: 5+ seconds with resolution control (480p/720p/1080p)
Open-source models with extended duration support.
Requires: request_id from original Seedance 2.0 generation
Optional: Continuation prompt
{
  "model": "seedance-v2.0-extend",
  "request_id": "original-video-id",
  "prompt": "Optional: describe what happens next",
  "duration": 10,
  "quality": "high"
}
The model seamlessly continues the original video, preserving style and motion. If no prompt is provided, it continues the scene naturally.

​
Example Prompts

A sweeping drone shot starts low over a misty forest at dawn, gradually rising above 
the tree canopy to reveal a vast mountain range in the distance. Morning light breaks 
through clouds, casting golden rays across the landscape. Smooth, cinematic camera 
movement with gradual ascent. Ultra-realistic, 4K quality, atmospheric fog.
Recommended Models: Kling v3.0 Pro, Veo 3.1, Seedance 2.0
A young woman in a red coat walks through a bustling Tokyo street at night, neon signs 
glowing around her. She stops to look up at the illuminated skyscrapers, her hair gently 
moving in the breeze. The camera follows her with a smooth tracking shot, capturing the 
vibrant city atmosphere. Cinematic, photorealistic, shallow depth of field.
Recommended Models: Wan 2.6, Kling v2.6 Pro, Sora 2 Pro
A sleek smartphone rotates slowly on a reflective surface, studio lighting highlighting 
its metallic edges. The screen lights up to show a colorful app interface. Camera orbits 
around the product with smooth 360-degree rotation. Clean, professional, high-end product 
commercial style, 4K quality.
Recommended Models: Runway Gen-3, Veo 3.1, Kling v2.6 Pro
A massive dragon perches on a cliff overlooking a medieval castle at sunset. Its scales 
shimmer in the golden light as it spreads its wings and roars, creating a gust of wind 
that ripples through the grass below. The camera slowly zooms in from a wide establishing 
shot. Epic fantasy, cinematic lighting, ultra-detailed, dramatic atmosphere.
Recommended Models: Pixverse v5.5, Seedance 2.0, Kling v3.0 Pro

​
Duration & Resolution Guide

Use CaseRecommended DurationResolutionModels
Social media clips5–8s720p–1080pSeedance 2.0, Wan 2.6, Kling v2.6 Pro
YouTube shorts10–15s1080pSora 2 Pro, Veo 3.1, Seedance 2.0
Product demos5–8s1080pRunway Gen-3, Veo 3.1, Kling v2.6 Pro
Cinematic sequences10–25s720p–1080pSora 2 Pro, Seedance 2.0
Quick prototypes4–6s480p–720pHailuo 2.3, Wan 2.2 Fast

​
API Endpoints

All text-to-video models use the unified endpoint pattern: 
POST /api/v1/{model-endpoint}
Refer to the Generate Video guide for integration details.

​
Model Selection Tips

For production work: Use Seedance 2.0, Kling v2.6 Pro, or Veo 3.1 for reliable quality.For long videos: Use Sora 2 Pro (25s), LTX 2 Fast (20s), or Seedance 2.0 (15s).For character animation: Use Wan 2.6 or Kling v2.6 Pro.For fast iteration: Use Seedance Lite, Wan 2.2 Fast, or Hailuo 2.3 Standard.For cinematic effects: Use Runway Gen-3 or Kling v3.0 Pro.For vertical content: Use models supporting 9:16 (most models support this).

​
Camera Movement Vocabulary

Include these terms in your prompt to guide camera motion:
Basic Movements:
  • Dolly In/Out — Move camera toward/away from subject
  • Pan Left/Right — Horizontal rotation
  • Tilt Up/Down — Vertical rotation
  • Zoom In/Out — Lens zoom (not camera movement)
Advanced Movements:
  • Crane Shot — Vertical movement (up/down)
  • Tracking Shot — Follow subject’s movement
  • Orbit/Arc — Circular movement around subject
  • Dutch Angle — Tilted camera for dramatic effect
Speeds:
  • Slow dolly — Smooth, cinematic
  • Crash zoom — Fast, dramatic zoom
  • Smooth tracking — Steadicam-like follow

​
Rate Limits

Video generation consumes significantly more credits than image generation through the Muapi.ai API. Cost factors include:
  • Duration: Longer videos cost more
  • Resolution: Higher resolutions cost more
  • Quality: High-quality modes cost more
  • Model: Premium models (Sora, Runway) cost more than open-source options
Refer to the Muapi.ai documentation for detailed pricing information.

​
Best Practices

  1. Start Short: Test with 5s generations before committing to longer durations.
  2. Be Specific: Describe camera movement, subject actions, and lighting in detail.
  3. Avoid Complex Edits: Text-to-video models work best with single continuous shots.
  4. Use Extend: For longer sequences, use Seedance 2.0 Extend to chain multiple clips.
  5. Resolution Trade-offs: Use 720p for iteration, 1080p for final output.

Build docs developers (and LLMs) love

Get started for freeTalk to us