ai-video-generator

Generate AI videos from text prompts using multiple providers (Runway, Pika, Sora). Includes script generation, scene planning, and batch processing.

SKILL.md

GitHub

AI Video Generator

Generate professional AI videos from text prompts using multiple providers.

Supported Providers

Provider Model Max Duration Quality
Runway Gen-3 10s High
Pika 1.0 4s Medium
Sora Turbo 20s Ultra

Usage

Basic Video Generation

from skills.ai_video_generator import VideoGenerator

gen = VideoGenerator(provider="runway")

# Simple prompt
video = gen.create(
    prompt="A cat walking in the rain at night, cinematic",
    duration=5,
    resolution="1080p"
)

# With scene planning
video = gen.create_from_script(
    script="""
    Scene 1: Wide shot of city skyline at sunset
    Scene 2: Close-up of person typing on laptop
    Scene 3: Product reveal with dramatic lighting
    """,
    style="cinematic"
)

Batch Processing

# Generate multiple videos
prompts = [
    "Product showcase: smartphone floating in space",
    "Tutorial: How to use the app",
    "Social media: 15-second promo clip"
]

videos = gen.batch_create(prompts, output_dir="./videos/")

Scene Planning

The skill automatically:

  1. Analyzes prompt for key elements
  2. Plans camera movements
  3. Suggests transitions
  4. Optimizes for target platform

Output Formats

  • MP4 — Standard video
  • GIF — Animated preview
  • WEBP — Animated thumbnail
  • Frames — Individual PNG frames

Configuration

# ~/.hermes/config/video-gen.yml
providers:
  runway:
    api_key: ${RUNWAY_API_KEY}
    default_style: cinematic
  pika:
    api_key: ${PIKA_API_KEY}
    default_style: anime

defaults:
  resolution: 1080p
  fps: 24
  output_dir: ~/videos/

Examples

Product Video

prompt: "Modern smartphone floating in zero gravity, 
        rotating slowly, dramatic lighting, 
        dark background with subtle particles"

Social Media Clip

prompt: "Quick montage of app features, 
        fast cuts, upbeat music vibe, 
        colorful gradients, 9:16 vertical"

Tutorial Video

prompt: "Screen recording style, 
        cursor clicking through UI, 
        text annotations appearing, 
        clean minimal design"

Pitfalls

  1. API Limits — Each provider has rate limits, use batch mode for multiple videos
  2. Duration Limits — Most providers limit to 4-10 seconds per generation
  3. Cost — Video generation can be expensive, preview with low-res first
  4. Aspect Ratio — Specify aspect ratio in prompt for consistent results