Midjourney API: Python SDK for V7, V8, and Niji

A small Python SDK for calling Midjourney V7, V8, and Niji through MuAPI. Generate four-image batches from text, optionally guide them with a reference image, and poll the asynchronous result with one API key.

▶ Watch: How to Access Midjourney API (V8, V7, Niji)

Related Projects

• Midjourney on MuAPI — model landing page, capabilities, and access details.
• Midjourney V8 playground — try the current image workflow in the browser.
• midjourney-comfyui — focused ComfyUI nodes for Midjourney V7, V8, and Niji.
• awesome-ai-image-models — compare image models by capability, API, and price.
• Open-Generative-AI — open-source studio for running generative image, video, and audio workflows.
• Flux-3-Dev-API — sibling Python SDK for unified FLUX image and video workflows.
• MiniMax-H3-API — sibling Python SDK for asynchronous generative-video jobs.
• Generative-Media-Skills — agent-ready skills for building generative-media pipelines.
• muapi-cli — CLI and MCP access to the same MuAPI model catalog.

Features

• Midjourney V7, V8, and Niji endpoint helpers
• Four images returned from each generation
• Optional image reference, aspect ratio, stylize, chaos, weird, negative prompt, and seed controls
• Asynchronous submit, poll, and webhook-ready requests
• Blocking wait_for_completion() and generate_and_wait() helpers
• Plain requests dependency and an injectable session for testing

Installation

pip install midjourney-api

Or install the latest source:

git clone https://github.com/Anil-matcha/midjourney-api.git
cd midjourney-api
pip install -e .

Set your MuAPI key:

export MUAPI_API_KEY=your_muapi_api_key

Create a key from the MuAPI access-key page.

Quick start

from midjourney_api import MidjourneyAPI

api = MidjourneyAPI()

task = api.v8(
    "A lone astronaut walking through a bioluminescent forest, cinematic lighting",
    aspect_ratio="16:9",
    stylize=500,
)

result = api.wait_for_completion(task["request_id"])
for image_url in api.extract_image_urls(result):
    print(image_url)

Use api.v7(...) or api.niji(...) for the other model variants. api.text_to_image(...) is a compatibility alias for V8.

Image references and controls

Pass a public image URL to guide the generation:

task = api.v7(
    "Editorial portrait with soft studio light and a deep red coat",
    image_url="https://example.com/reference.jpg",
    aspect_ratio="4:3",
    stylize=250,
    chaos=10,
    weird=0,
    negative_prompt="text, watermark, blurry",
)

stylize ranges from 0–1000, chaos from 0–100, and weird from 0–3000. Supported aspect ratios are 1:1, 16:9, 9:16, 3:4, 4:3, and 21:9.

API endpoints

Workflow	Endpoint
Midjourney V7	POST /api/v1/midjourney-v7
Midjourney V8	POST /api/v1/midjourney-v8
Midjourney Niji	POST /api/v1/midjourney-niji
Poll task	GET /api/v1/predictions/{request_id}/result

All requests use the x-api-key header. The generation endpoints are asynchronous and return a request_id immediately.

Raw request shape

curl --location --request POST "https://api.muapi.ai/api/v1/midjourney-v8" \
  --header "x-api-key: $MUAPI_API_KEY" \
  --header "Content-Type: application/json" \
  --data-raw '{
    "prompt": "A celestial observatory above a waterfall at night",
    "aspect_ratio": "16:9",
    "stylize": 500,
    "chaos": 10,
    "webhook_url": "https://example.com/webhook"
  }'

The SDK sends webhook_url in the JSON body when supplied. MuAPI posts the completed result to that URL while the polling endpoint remains available.

Development

python -m unittest discover -s tests -v

The tests mock HTTP calls and never spend credits.

License

MIT — see LICENSE.