@muapi/ai-sdk-provider

MuAPI provider for the Vercel AI SDK — image and video generation with 400+ models including Flux, Veo3, Kling, Wan, Seedance, Midjourney, and more.

Installation

npm install @muapi/ai-sdk-provider
# or
pnpm add @muapi/ai-sdk-provider
# or
yarn add @muapi/ai-sdk-provider

Setup

Get your API key from muapi.ai/dashboard/api-keys.

import { createMuapi } from '@muapi/ai-sdk-provider';

const muapi = createMuapi({
  apiKey: process.env.MUAPI_API_KEY,
});

Or use the default singleton (reads MUAPI_API_KEY from env):

import { muapi } from '@muapi/ai-sdk-provider';

Image Generation

import { muapi } from '@muapi/ai-sdk-provider';
import { generateImage } from 'ai';

const { images } = await generateImage({
  model: muapi.image('flux-schnell'),
  prompt: 'a sunset over the ocean, photorealistic',
});

console.log(images[0].url); // output URL

Available image models

Model ID	Description
flux-schnell	Flux Schnell — fast, high quality
flux-dev	Flux Dev
flux-kontext-dev / flux-kontext-pro / flux-kontext-max	Flux Kontext (text-to-image)
hidream-fast / hidream-dev / hidream-full	HiDream
midjourney / midjourney-v7 / midjourney-v8	Midjourney
gpt4o	GPT-4o image generation
gpt-image-2	GPT Image 2
imagen4 / imagen4-fast / imagen4-ultra	Google Imagen 4
seedream / seedream-5	Seedream
wan2.1 / wan2.5 / wan2.6	Wan
qwen / qwen-2 / qwen-2-pro	Qwen
reve	Reve
ideogram	Ideogram v3
hunyuan	Hunyuan

Video Generation

import { muapi } from '@muapi/ai-sdk-provider';
import { experimental_generateVideo as generateVideo } from 'ai';

const { videos } = await generateVideo({
  model: muapi.video('veo3-fast'),
  prompt: 'a timelapse of clouds moving over mountains',
});

console.log(videos[0].url); // MP4 URL

Available video models

Model ID	Description
veo3 / veo3-fast / veo3.1 / veo4	Google Veo
kling-master / kling-v2.5-pro / kling-v3-pro	Kling
wan2.1 / wan2.2 / wan2.5 / wan2.6 / wan2.7	Wan
seedance-pro / seedance-pro-fast	Seedance
runway	Runway
pixverse / pixverse-v5 / pixverse-v6	Pixverse
sora / sora-2	OpenAI Sora
hunyuan	Hunyuan
vidu / vidu-q2-pro / vidu-q3-pro	Vidu

Image-to-Video

const { videos } = await generateVideo({
  model: muapi.imageToVideo('kling-master'),
  prompt: 'zoom in slowly',
  providerOptions: {
    muapi: { image_url: 'https://example.com/image.jpg' },
  },
});

Provider options

Pass extra model-specific parameters via providerOptions.muapi:

const { images } = await generateImage({
  model: muapi.image('flux-dev'),
  prompt: 'a cat',
  providerOptions: {
    muapi: {
      width: 1024,
      height: 768,
      num_inference_steps: 28,
    },
  },
});

License

MIT