Automatic1111 | AI SDK

AUTOMATIC1111 is a popular web interface for Stable Diffusion that provides a comprehensive set of features for image generation. The Automatic1111 provider for the AI SDK enables seamless integration with locally hosted AUTOMATIC1111 instances while offering unique advantages:

Local Control: Full control over your image generation with local Stable Diffusion models
No API Costs: Generate unlimited images without per-request charges
Model Flexibility: Use any Stable Diffusion checkpoint
Privacy: All generation happens locally on your hardware
Community Models: Access to thousands of community-created models from Civitai and HuggingFace

Learn more about AUTOMATIC1111's capabilities in the AUTOMATIC1111 Documentation.

Setup

You need to have AUTOMATIC1111 running with the --api flag enabled. Start your AUTOMATIC1111 instance with:

# Windows
./webui.bat --api

# Linux/Mac
./webui.sh --api

The Automatic1111 provider is available in the automatic1111-provider module. You can install it with:

# pnpm
pnpm add automatic1111-provider

# npm
npm install automatic1111-provider

# yarn
npm install automatic1111-provider

Provider Instance

To create an Automatic1111 provider instance, use the createAutomatic1111 function:

import { createAutomatic1111 } from 'automatic1111-provider';

const automatic1111 = createAutomatic1111({
  baseURL: 'http://127.0.0.1:7860', // Your AUTOMATIC1111 instance
});

Image Models

The Automatic1111 provider supports image generation through the image() method:

// Basic image generation
const imageModel = automatic1111.image('v1-5-pruned-emaonly');

// With custom model
const sdxlModel = automatic1111.image('sd_xl_base_1.0');

Examples

Basic Image Generation

import { automatic1111 } from 'automatic1111-provider';
import { experimental_generateImage as generateImage } from 'ai';

const { images } = await generateImage({
  model: automatic1111.image('v1-5-pruned-emaonly'),
  prompt: 'A beautiful sunset over mountains',
  size: '512x512',
});

Advanced Configuration

const { images } = await generateImage({
  model: automatic1111.image('realistic-vision-v4'),
  prompt: 'Portrait of a wise old wizard with a long beard',
  n: 2,
  seed: 12345,
  providerOptions: {
    automatic1111: {
      negative_prompt: 'blurry, ugly, deformed, low quality',
      steps: 40,
      cfg_scale: 8.5,
      sampler_name: 'DPM++ SDE Karras',
      styles: ['photorealistic', 'detailed'],
      check_model_exists: true,
    },
  },
});

Provider Options

The Automatic1111 provider supports the following options for customizing image generation:

Available Options

Option	Type	Default	Description
`negative_prompt`	`string`	`undefined`	What you don't want in the image
`steps`	`number`	`20`	Number of sampling steps
`cfg_scale`	`number`	`7`	CFG (Classifier Free Guidance) scale
`sampler_name`	`string`	`"Euler a"`	Sampling method
`denoising_strength`	`number`	`undefined`	Denoising strength for img2img (0.0-1.0)
`styles`	`string[]`	`undefined`	Apply predefined styles
`check_model_exists`	`boolean`	`false`	Verify model exists before generation

Model Management

The provider automatically detects available models from your AUTOMATIC1111 instance. To use a model:

Place your .safetensors or .ckpt model files in the models/Stable-diffusion/ folder
Restart AUTOMATIC1111 or refresh the models list in the web interface
Use the exact model name (without file extension) in the provider