Magic Hour Tool

Magic Hour is an AI video and image generation API. One API key gives access to Sora 2, Veo 3.1, Kling 3.0, Seedance, MiniMax H3, WAN 2.2 and LTX 2.3 for video, and GPT-image, Nano Banana Pro, Seedream, Flux and Z-Image for images.

Get a free API key at the Magic Hour developer page (400 credits on signup + 100/day, no card required) and export it as MAGIC_HOUR_API_KEY.

Installation

pip install git+https://github.com/RhythmP28/llama-index-tools-magic-hour  # until it lands on PyPI

Usage

This tool has a more extensive example usage documented in a Jupyter notebook here.

from llama_index.tools.magic_hour import MagicHourToolSpec
from llama_index.core.agent.workflow import FunctionAgent
from llama_index.llms.openai import OpenAI

magic_hour_tool = MagicHourToolSpec(
    api_key="your-key"
)  # or MAGIC_HOUR_API_KEY
agent = FunctionAgent(
    tools=magic_hour_tool.to_tool_list(),
    llm=OpenAI(model="gpt-4o"),
)

await agent.run("Make a 5 second 480p video of a corgi surfing at sunset")

Calling the spec directly:

result = magic_hour_tool.text_to_video("a corgi surfing at sunset")
# {'project_id': 'cuid...', 'status': 'complete', 'credits_charged': 120,
#  'video_url': 'https://...'}

Available Functions

All functions have sync and async (a-prefixed) variants and return a plain dict.

text_to_video(prompt, model="wan-2.2", duration_seconds=5, resolution="480p", aspect_ratio="16:9"): Generate a video from text. Returns project_id, status, video_url, credits_charged.

image_to_video(image_url_or_path, prompt, model="wan-2.2", duration_seconds=5, resolution="480p"): Animate an image (public URL or local path) into a video. Same return shape.

generate_image(prompt, model="default", image_count=1, aspect_ratio="16:9"): Generate images from text. Returns project_id, status, image_urls, credits_charged.

Constructor options: api_key, wait_for_completion (default True, polls until the job finishes), download_outputs (default False) and download_directory. Failed jobs are refunded automatically.

Models

Image models (about 5 credits per image): default, gpt-image-2, nano-banana-pro, seedream-5-pro, flux-2-klein, z-image-turbo, qwen-edit.

Links

• Magic Hour API docs
• Python SDK magic_hour
• Other integrations: Vercel AI SDK provider, ComfyUI node, MCP server

This loader is designed to be used as a way to load data as a Tool in an Agent.