# Stable Diffusion XL - Core.Today AI API

> Stability AI's classic SDXL (85M+ runs) — the battle-tested text-to-image model with img2img, inpainting, refiner, and LoRA support. A dependable workhorse with a huge ecosystem.

- **Provider**: Stability AI
- **Model ID**: stability-ai/sdxl
- **Category**: Image Generation
- **Credits**: 9 per image (num_outputs billed per image)
- **Speed**: Medium
- **Quality**: High

## Features
- Battle-tested classic (85M+ runs)
- img2img and inpainting with mask
- Expert-ensemble refiner support
- LoRA weights loading
- Up to 4 images per request

## Use Cases
- General-purpose image generation
- Inpainting and image editing
- LoRA-based custom styles
- Compatibility with SDXL ecosystem workflows
- Cost-efficient production pipelines

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Optional
- **prompt**: string (default: An astronaut riding a rainbow unicorn) - Input prompt
- **negative_prompt**: string - What to avoid in the image
- **image**: string - Input image for img2img or inpainting
- **mask**: string - Inpainting mask — white areas are repainted, black preserved
- **width**: integer (default: 1024) - Output width
- **height**: integer (default: 1024) - Output height
- **num_outputs**: integer (default: 1) - Number of images (1-4). Billed per image
- **scheduler**: string (default: K_EULER) - scheduler
  Options: DDIM, DPMSolverMultistep, HeunDiscrete, KarrasDPM, K_EULER_ANCESTRAL, K_EULER, PNDM
- **num_inference_steps**: integer (default: 50) - Denoising steps (fewer is faster)
- **guidance_scale**: number (default: 7.5) - Classifier-free guidance scale
- **prompt_strength**: number (default: 0.8) - img2img/inpaint denoising strength. 1.0 fully replaces the input
- **seed**: integer - Random seed. Leave blank to randomize
- **refine**: string (default: no_refiner) - Refine style: no_refiner, expert_ensemble_refiner, base_image_refiner
  Options: no_refiner, expert_ensemble_refiner, base_image_refiner
- **high_noise_frac**: number (default: 0.8) - For expert_ensemble_refiner, the fraction of noise to use
- **refine_steps**: integer - For base_image_refiner, the number of steps to refine, defaults to num_inference_steps
- **apply_watermark**: boolean (default: true) - Applies a watermark to enable determining if an image is generated in downstream applications. If you have other provisions for generating or deploying images safely, you can use this to disable watermarking.
- **lora_scale**: number (default: 0.6) - LoRA additive scale. Only applicable on trained models.
- **replicate_weights**: string - Replicate LoRA weights to use. Leave blank to use the default weights.
- **disable_safety_checker**: boolean (default: false) - Disable safety checker for generated images. This feature is only available through the API. See [https://replicate.com/docs/how-does-replicate-work#safety](https://replicate.com/docs/how-does-replicate-work#safety)

## Examples

### Classic Text-to-Image
Generate a 1024px image with the refiner
```json
{
  "model": "stability-ai/sdxl",
  "input": {
    "prompt": "An astronaut riding a rainbow unicorn, cinematic, dramatic",
    "width": 1024,
    "height": 1024,
    "refine": "expert_ensemble_refiner"
  }
}
```

### Inpainting
Repaint the masked region of an image
```json
{
  "model": "stability-ai/sdxl",
  "input": {
    "prompt": "A golden retriever sitting on a bench",
    "image": "https://example.com/photo.png",
    "mask": "https://example.com/mask.png",
    "prompt_strength": 0.9
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- Use expert_ensemble_refiner for noticeably better detail on final renders
- For inpainting, white mask areas are repainted and black areas preserved
- Lower num_inference_steps (25-30) for faster drafts with minor quality loss
- For instant drafts use SDXL Lightning; for frontier quality use FLUX or Seedream

## Documentation
https://replicate.com/stability-ai/sdxl