NewReference images & transparent backgrounds

Chat, image and video AI
through one API key

Three models ready for your product — muse-chat, muse-image and muse-video. OpenAI-compatible: keep the SDK you already use and just change base_url.

OpenAI SDK for Python & Node.js Streaming SSE Pay as you go
from openai import OpenAI

client = OpenAI(
    base_url="https://muse.isemi.io/v1",
    api_key="m2a-your-key",
)

img = client.images.generate(
    model="muse-image",
    prompt="A red fox in an autumn forest, morning light",
    size="16:9",
)
print(img.data[0].url)

Change base_url and go

Same request and response format as OpenAI. Works with the OpenAI SDKs, LobeChat, Cherry Studio, ChatGPT-Next-Web and more.

Real-time streaming

Chat replies arrive chunk by chunk over Server-Sent Events, so you can show the answer while it is being written.

Async tasks

Images and videos run in the background: create a task, get an ID, poll for progress. No proxy or client timeouts.

Reference images

Send up to 4 images to keep a character, product or layout. Transparent PNG output when you need it.

Models

Three models for chat, images and video

Pick a model with the model parameter, just like with OpenAI. One API key works for all of them.

Chat

muse-chat

A multi-turn assistant that also understands images in your messages.

  • Streaming and non-streaming
  • System prompts and conversation history
  • Image input (vision)
POST/v1/chat/completions
Image

muse-image

Generate images from a text prompt, or from reference images you attach.

  • Aspect ratio of your choice: 1:1, 16:9, 9:16…
  • Up to 4 reference images per request
  • Transparent PNG when you need it
  • Returned as a URL or base64
POST/v1/images/generationsPOST/v1/images/edits
Video

muse-video

Generate video from a text prompt, or from an image used as the first frame.

  • Text-to-video and image-to-video
  • Choose duration and aspect ratio
  • Runs as a task with progress tracking
POST/v1/videos
To plug into existing tools quickly, familiar model names such as gpt-4o, dall-e-3 or sora are also accepted and routed to the matching Muse model. See the full list in the docs.
Pricing

Simple, pay-as-you-go pricing

Top up credit once and use it across every model. Each successful request is deducted from your balance — failed requests are never charged.

Video

muse-video

$0.006/ second
$0.06 for a 10-second video
  • Text-to-video and image-to-video
  • Billed on the duration you request
  • Progress tracking for every task
Top up credit

Chat

muse-chat

$1/ 1M input tokens
$3 / 1M output tokens
  • Streaming and vision
  • Token usage and cost in every response
  • Multi-turn conversations
Top up credit
$10 of credit ≈
666images
16610-second videos
10Minput tokens of chat

Prices in USD. Check your balance at any time with GET /v1/balance; every response includes its cost_usd.

Get started

Up and running in minutes

Choose an amount

Open Buy credit, pick how much credit you want (from $5) and enter your email.

Pay with PayPal or card

Pay with PayPal or a debit or credit card. Your API key (m2a-…) appears instantly with the credit loaded. It is shown only once, so store it safely.

Start calling the API

Set base_url to https://muse.isemi.io/v1 and make your first request. The docs have complete examples.

FAQ

Frequently asked questions

Do I need to change my code?

No. If you already use an OpenAI SDK or an app that supports an OpenAI-compatible endpoint, set the base URL to https://muse.isemi.io/v1 and swap in your API key. Requests and responses keep the same format.

How is usage billed?

Images cost $0.015 each, video $0.006 per requested second, and chat $1 per 1M input tokens and $3 per 1M output tokens. You are charged only when a request succeeds: if a generation fails, the reserved amount is refunded automatically.

How do I check my balance, and what happens when it runs out?

Call GET /v1/balance with your key at any time. When the balance is too low for a request, the API answers HTTP 402 with insufficient_balance until you top up — nothing is charged.

How long does an image or video take?

An image usually takes 1–2 minutes and a video longer; it can be slower at peak times. That is why we recommend async tasks ("async": true for images, always on for video) and polling for the result instead of holding one connection open.

How long are generated files kept?

Results are returned as random, hard-to-guess URLs. Download and store them on your side right away — links are not guaranteed to last. For images you can also get base64 directly with "response_format": "b64_json".

Which chat apps work with it?

Any app that lets you configure an OpenAI-compatible endpoint: LobeChat, Cherry Studio, ChatGPT-Next-Web, Open WebUI… Enter the base URL and your API key. See Integrations in the docs.

Can my team use several keys?

Yes. Ask us for one key per app or teammate; each key has its own balance and can be revoked on its own.

Get an API key

Ready to build?

Buy credit online and get your API key instantly, or email us about volume pricing and invoices.