sora-video

name	sora-video
description	Generates AI videos from text prompts or reference images using OpenAI Sora. Use when the user wants to create videos, generate video clips, make AI films, animate images, or produce motion content. Supports text-to-video, image-to-video, and video remixing.

Sora Video Generation

Generate high-quality AI videos using OpenAI's Sora model.

Capabilities

• Text-to-video: Create video clips from text descriptions
• Image-to-video: Animate a reference image into a video clip
• Remix: Create variations of previously generated videos

Requirements

Before using this skill, ensure:

1. Python packages are installed:

   pip install openai
   

2. API key is set:

   export OPENAI_API_KEY="your-openai-api-key"
   

   Get your key at: https://platform.openai.com/api-keys

Usage

Text-to-Video

Generate a video from a text description:

python skills/sora-video/scripts/generate_video.py --prompt "A drone shot flying over a misty forest at sunrise" --output forest.mp4

Image-to-Video

Animate a reference image:

python skills/sora-video/scripts/generate_video.py --prompt "The scene comes to life with gentle wind" --input photo.jpg --output animated.mp4

Higher Quality (Pro)

Use the pro model for production-grade quality:

python skills/sora-video/scripts/generate_video.py --prompt "A cinematic sunset over mountains" --model sora-2-pro --size 1792x1024 --seconds 12 --output cinematic.mp4

Script Arguments

Argument	Short	Required	Default	Description
--prompt	-p	Yes	--	Text description of the desired video
--output	-o	No	generated_video.mp4	Output file path
--input	-i	No	--	Reference image for image-to-video mode
--model	-m	No	sora-2	Model: sora-2 or sora-2-pro
--seconds	-s	No	8	Duration: 4, 8, or 12
--size		No	1280x720	Resolution (see table below)
--timeout		No	600	Max wait time in seconds

Available Sizes

Size	Orientation	Models
1280x720	Landscape	sora-2, sora-2-pro
720x1280	Portrait	sora-2, sora-2-pro
1792x1024	Landscape HD	sora-2-pro only
1024x1792	Portrait HD	sora-2-pro only

Prompt Tips

For best results:

1. Be cinematic: Describe camera movement, lighting, and atmosphere
2. Specify style: "photorealistic", "anime style", "watercolor animation"
3. Describe motion: "slowly panning", "time-lapse", "tracking shot"
4. Set the mood: "dramatic lighting", "golden hour", "neon-lit"

Example Prompts

Nature:

A drone shot flying over a misty forest at sunrise, golden light filtering through
the trees, birds rising from the canopy, cinematic quality

Urban:

A time-lapse of a busy Tokyo intersection at night, neon signs reflecting on wet
pavement, crowds crossing in all directions, cyberpunk atmosphere

Abstract:

Liquid metal flowing and morphing into geometric shapes, iridescent reflections,
dark background, smooth motion, high-end product visualization style

Pricing

Model	Resolution	Per Second	4s Clip	8s Clip	12s Clip
sora-2	720p	$0.10	$0.40	$0.80	$1.20
sora-2-pro	720p	$0.30	$1.20	$2.40	$3.60
sora-2-pro	1024p	$0.50	$2.00	$4.00	$6.00

Note: You are charged for every generation attempt, including failed ones.

Workflow

1. Understand what the user wants to create
2. Craft a detailed, cinematic prompt (camera angles, lighting, mood, motion)
3. Choose appropriate model, duration, and resolution
4. Run the generation script
5. If the user provides an image, use image-to-video mode with --input
6. Present the generated video to the user
7. Offer to iterate or adjust based on feedback

Troubleshooting

"OPENAI_API_KEY not set"

• Set the environment variable: export OPENAI_API_KEY="your-key"

"openai package not installed"

• Install it: pip install openai

"Video generation failed: content_policy_violation"

• The prompt was flagged by content moderation
• Remove references to real people, copyrighted characters, or inappropriate content
• Rephrase the prompt and try again

"Video generation timed out"

• Pro model and higher resolutions take longer (up to 5 minutes)
• Increase timeout: --timeout 900
• Try with sora-2 (faster) instead of sora-2-pro

"Rate limit exceeded"

• Wait a moment and try again
• Standard API tier allows ~10 requests per minute

Output

The script will:

1. Print progress messages with status and percentage
2. Save the MP4 video to the specified output path
3. Print "SUCCESS" with the file path on success
4. Return exit code 0 on success, 1 on failure