Magic Hour gives Openclaw Skills a unified API for generating AI videos and images across leading models with asynchronous job tracking and optional local downloads.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install magic-hour
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install magic-hour using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Magic Hour?
Magic Hour is an AI media-generation skill that connects agents to video and image models including Sora 2, Veo 3.1, Kling 3.0, WAN 2.2, LTX 2.3, GPT-image, Nano Banana Pro, Flux, and Z-Image. It supports text-to-video, image-to-video, and text-to-image workflows through one API key, making model selection and content production easier to automate.
The skill supports both a hosted MCP server with zero local installation and a Python SDK script workflow. Openclaw Skills users can submit asynchronous jobs, monitor project status, retrieve expiring output URLs, download generated assets, estimate credit usage, and handle failures through structured JSON responses.
Magic Hour Use Cases
- Create cinematic AI video clips from natural-language prompts for campaigns, presentations, prototypes, and social media.
- Animate a local image or public image URL with controlled camera movement, subject motion, and visual effects.
- Generate product concepts, illustrations, social tiles, photorealistic scenes, and edited images.
- Produce vertical 9:16 shorts, landscape 16:9 videos, or square 1:1 social content.
- Compare cost, duration, resolution, and quality across multiple video models through one interface.
- Queue long-running generation tasks and check their status later without blocking the agent.
- Use Openclaw Skills with an MCP-compatible client when a zero-install hosted integration is preferred.
How Magic Hour Works
- Configure the Magic Hour API key through the
MAGIC_HOUR_API_KEYenvironment variable, or attach the hosted MCP server with bearer authentication. - Select a generation workflow: text-to-video, image-to-video, or text-to-image.
- Choose a compatible model, duration, resolution, aspect ratio, and optional audio setting.
- Submit the request. The API creates an asynchronous project and returns a project identifier.
- The scripts poll approximately every five seconds when waiting is enabled, or return immediately with
status: "queued"when--no-waitis used. - Read the structured JSON result, including project ID, status, model, output URL or URLs, and credits charged.
- Optionally use
--download-dirto save generated files locally before temporary output URLs expire. - If a job fails, inspect the structured error, correct common issues such as unsupported duration or resolution, and retry once; failed jobs are automatically refunded.
Magic Hour Setup
Option 1: Hosted MCP server
Use the remote MCP endpoint at https://mcp.magichour.ai/ when your agent supports MCP servers. Configure the Authorization: Bearer $MAGIC_HOUR_API_KEY header. Claude web and desktop use OAuth with the Magic Hour MCP client ID.
Option 2: Python script workflow
- Create a free Magic Hour developer account and API key at
https://magichour.ai/developer. New accounts receive 400 signup credits and 100 credits per day without a card. - Export the API key and install the official Python SDK:
export MAGIC_HOUR_API_KEY=mhk_...
pip install magic_hour
Python 3.9 or newer is required. Run the included scripts from the skill directory:
python3 {baseDir}/scripts/text_to_video.py "a corgi surfing at golden hour, slow-motion, cinematic" \
--model wan-2.2 --duration 5 --resolution 480p --aspect-ratio 16:9 --download-dir ./out
python3 {baseDir}/scripts/image_to_video.py ./photo.png "slow push-in, hair moves in the wind" \
--model kling-3.0 --duration 5 --resolution 720p
python3 {baseDir}/scripts/generate_image.py "isometric cozy coffee shop, soft morning light" \
--model nano-banana-pro --count 2 --aspect-ratio 1:1
python3 {baseDir}/scripts/status.py <project_id> --kind video --wait
Use --no-wait for immediate queue submission. Review the model catalogue and raw HTTP examples in the skill's references/models.md and references/api.md files when using unsupported environments or direct API requests.
Magic Hour Data Schema & Taxonomy
Response object
Every script prints one JSON object to standard output. Typical fields include:
| Field | Purpose |
|---|---|
project_id |
Identifier used to poll or retrieve a generation project. |
status |
Job state such as queued, completed, or error. |
model |
Magic Hour model used for the request. |
url / urls |
Temporary URL or URLs for generated media. |
downloaded_paths |
Local file paths when --download-dir is supplied. |
credits_charged |
Credits consumed by the completed request. |
error |
Structured failure details, including an error message when applicable. |
Asset and metadata organization
- Video projects are tracked by
project_idand queried withstatus.py --kind video. - Image generation can return multiple outputs through the
urlscollection. - Local files are written beneath the directory supplied to
--download-dir. - Source images may be local files, which are uploaded automatically, or public HTTPS URLs.
- Model metadata includes supported duration, credits per second, resolution constraints, and aspect-ratio behavior.
- Cost is calculated as credits per second multiplied by video duration; image generation generally costs about 5 to 50 credits per image.
- Failures return a non-zero process status and a JSON payload shaped like
{"status":"error","error":{...}}.
Magic Hour Advanced Features
- Hosted MCP integration exposes image, video, and audio generation tools without a local SDK installation.
- MCP tooling supports asset uploads through
videoAssets_generatePresignedUrland dedicated wait/status operations. - Multi-model generation covers free and premium options, from WAN 2.2, LTX 2.3, and MiniMax H3 to Sora 2 and Veo 3.1.
- Asynchronous execution supports queued jobs, polling, delayed status checks, and non-blocking
--no-waitworkflows. - Automatic local downloading protects assets from expiring remote URLs.
- Image-to-video workflows automatically upload local source images and preserve the source image as the visual foundation.
- Video generation supports model-specific duration validation, resolution controls, aspect ratios, and optional Veo audio.
- Structured JSON output makes the skill easy to connect to pipelines, automation scripts, and other Openclaw Skills.
- Cost-aware operation helps agents prefer free models and notify users before spending more than approximately 500 credits.
SKILL.md
Loading
Related Openclaw Skills
OpenClaw Agent Cognitive OS X
A cognitive control layer that turns OpenClaw goals into adaptive, verified, and safely managed autonomous-agent workflows.
OpenClaw Plugin Intelligence
OpenClaw Plugin Intelligence helps agents identify, select, securely execute, validate, and recover from external plugin workflows.
Icon Designer Generator
Generate editable AI icon designs, posters, and multi-page visual carousels from text prompts, templates, and customizable design specifications.
AI Photo Editor
AI Photo Editor generates editable posters, photo assets, and multi-page image carousels from text prompts, templates, and configurable design layers.
Academic Paper Reviewer
A structured five-seat AI review panel that evaluates academic manuscripts from journal-fit, methodology, domain, cross-disciplinary, and Devil's Advocate perspectives.
Submit to WorthToTry
Automate AI tool submissions to WorthToTry by auditing a product page, drafting its listing, and returning a review link.
Featured*
