Generate Video

​

Overview

• Text-to-video - Generate videos from natural language descriptions
• Image-to-video - Animate static images into dynamic videos
• Multiple resolutions - From social media to high-definition formats
• Consistent motion - Smooth, natural movement and transitions
• Temporal coherence - Maintains visual consistency across frames

• Social media content - Create eye-catching videos for posts and Stories
• Marketing campaigns - Generate product demos and promotional videos
• Content creation - B-roll footage for videos and presentations
• E-commerce - Product showcases with dynamic movement
• Prototyping - Visualize concepts before expensive production
• Animation - Quick animated sequences for various purposes
• Education - Visual demonstrations and explainer content

​

How Video Generation Works

1. Prompt/Image Analysis - Understanding what needs to be generated and how it should move
2. Motion Planning - Determining camera movement, object motion, and scene dynamics
3. Frame Generation - Creating each frame with temporal consistency
4. Motion Smoothing - Ensuring smooth transitions between frames
5. Upscaling & Enhancement - Improving quality and resolution
6. Encoding - Compressing into MP4 format
7. Delivery - Encoded as base64 data URI for transmission

​

Writing Effective Video Prompts

​

Prompt Structure for Videos

[Subject] [action/motion] [environment] [camera movement] [style/mood]

• ✅ “Ocean waves crashing against rocky cliffs, slow motion, cinematic”
• ✅ “Golden retriever running through meadow, camera following, sunny day”
• ✅ “City traffic at night, time lapse, neon lights, aerial view”
• ✅ “Waterfall cascading down mountain, morning mist, drone footage”

• ❌ “Beach” (no motion described)
• ❌ “Person standing still” (minimal movement)
• ❌ “Multiple complex scenes” (too ambitious for 8 seconds)

​

Key Elements for Video Prompts

• Natural motion: “leaves rustling in wind,” “water flowing”
• Object motion: “car driving,” “bird flying,” “person walking”
• Camera motion: “camera panning,” “zooming in,” “orbiting around”

• Set the scene: “in a bustling city,” “on a quiet beach,” “through a forest”
• Include atmosphere: “foggy morning,” “golden hour,” “stormy weather”

• “Slow motion,” “time lapse,” “steady cam,” “drone footage”
• “Close-up,” “wide shot,” “tracking shot,” “aerial view”
• “Handheld,” “smooth pan,” “dolly zoom”

• “Cinematic,” “documentary style,” “dreamy,” “dramatic”
• “Realistic,” “artistic,” “commercial quality”

​

Video-Specific Tips

• ✅ Describe movement clearly
• ✅ Keep it simple - one main action or scene
• ✅ Specify camera movement if important
• ✅ Use temporal terms (slow motion, time lapse)
• ✅ Focus on visually interesting subjects

• ❌ Try to fit multiple scenes in 8 seconds
• ❌ Describe complex narratives
• ❌ Request rapid scene changes
• ❌ Include dialogue or specific audio
• ❌ Expect photorealistic human faces/actions (challenging for AI)

​

Examples

from incredible_python import Incredible
import base64

client = Incredible(api_key="YOUR_API_KEY")

response = client.generate_video(
    prompt="Ocean waves",
    size="1280x720"
)

print(f"Video ID: {response.video_id}")
print(f"Duration: {response.duration}s")

# Extract and save video
video_data = response.video_url.split(',')[1]
with open('generated_video.mp4', 'wb') as f:
    f.write(base64.b64decode(video_data))

print("Video saved!")

​

Request Parameters

​

prompt (required)

​

size (optional)

• Aspect ratio: 16:9
• Best for: YouTube, presentations, website embeds
• File size: Moderate
• Generation time: Standard

• Aspect ratio: 9:16
• Best for: Instagram/TikTok Stories, mobile-first content
• File size: Moderate
• Generation time: Standard

• Aspect ratio: 16:9
• Best for: High-quality YouTube, professional content
• File size: Larger
• Generation time: Longer

• Aspect ratio: 9:16
• Best for: Premium mobile content, vertical video platforms
• File size: Larger
• Generation time: Longer

• Aspect ratio: 1:1
• Best for: Instagram posts, social feeds
• File size: Moderate
• Generation time: Standard

• Mobile/Social: Use portrait (9:16) or square (1:1)
• YouTube/Web: Use landscape (16:9)
• Quality vs Speed: Lower resolutions generate faster
• File size: Higher resolutions = larger files

​

input_reference (optional)

• Animate photos - Bring static images to life
• Product demonstrations - Show products in motion
• Brand consistency - Start from your existing visual assets
• Character animation - Animate illustrations or artwork
• Architectural visualization - Add life to renderings

import base64

# Load your image
with open("reference.jpg", "rb") as f:
    image_data = f.read()

# Encode as base64
encoded_image = base64.b64encode(image_data).decode('utf-8')

# Generate video from image
response = client.generate_video(
    prompt="Camera slowly pans across the scene, gentle movement",
    input_reference=encoded_image,
    size="1280x720"
)

• Start with high-quality, well-composed images
• Describe the motion/animation you want
• Keep motion expectations realistic (subtle movements work best)
• Avoid images with too much complexity

​

Best Practices

• Focus on a single, clear action or scene
• Describe motion explicitly (“waves crashing,” “camera panning”)
• Include camera movement for more dynamic results
• Use cinematic terms for professional looks
• Specify pace (slow motion, time lapse)

• Lower resolutions generate faster and use less bandwidth
• Portrait formats are ideal for mobile platforms
• Landscape formats for traditional viewing
• Square for universal social media compatibility

• Video generation takes 1-10 minutes - implement proper timeout handling
• Consider queueing or background processing for user-facing apps
• Cache frequently requested videos
• Provide loading indicators and progress updates to users

• Use clear, simple prompts
• Avoid overly complex scenes
• Natural phenomena work well (water, fire, clouds)
• Camera movements are easier than complex object actions
• Test with various prompts to learn what works best

• Async processing - Start generation, notify user when complete
• Preview thumbnails - Generate image first to preview before video
• Batch generation - Queue multiple videos for efficient processing
• Fallback strategy - Have backup content if generation fails

​

Handling Video Output

data:video/mp4;base64,AAAIGZ0...

# Python
import base64

# Extract base64 data (remove data URI prefix)
video_data = response.video_url.split(',')[1]

# Decode and save
with open('video.mp4', 'wb') as f:
    f.write(base64.b64decode(video_data))

// TypeScript/Node.js
import fs from 'fs';

// Extract base64 data
const base64Data = response.video_url.split(',')[1];

// Decode and save
const videoBuffer = Buffer.from(base64Data, 'base64');
fs.writeFileSync('video.mp4', videoBuffer);

// Browser JavaScript
// Create download link
const link = document.createElement('a');
link.href = response.video_url;
link.download = 'generated_video.mp4';
link.click();

• Ensures reliable transmission over HTTP
• No need for separate file hosting
• Immediate availability
• Simplified API response structure

• 720p: ~1-3MB for 8 seconds
• 1080p: ~2-5MB for 8 seconds
• Depends on complexity and motion

​

Important Notes

• Resolution requested
• Scene complexity
• Current server load
• Whether using image-to-video

• Video length: Fixed at 8 seconds
• No audio generation (silent videos)
• Best results with natural phenomena and simple motions
• Complex human actions may not be photorealistic

​

Response

{
  "success": true,
  "video_id": "operation_abc123",
  "video_url": "data:video/mp4;base64,AAAIGZ0...",
  "status": "completed",
  "duration": 8
}