> ## Documentation Index
> Fetch the complete documentation index at: https://docs.firemoon.studio/llms.txt
> Use this file to discover all available pages before exploring further.

# FLUX Provider

> Fast, high-quality image generation with FLUX models

## Overview

FLUX is optimized for speed and creative control. It delivers sub-second inference times while maintaining high image quality.

## Available Models

### FLUX/dev

The primary FLUX model optimized for both speed and quality.

**Endpoint:** `POST /api/v1/flux/dev`

**Best for:**

* Rapid prototyping
* Content creation workflows
* Real-time applications
* High-volume generation

**Performance:**

* **Inference time:** \~0.5-2 seconds
* **Quality:** High
* **Cost:** Low

## Parameters

### Required Parameters

| Parameter | Type   | Description                                                     |
| --------- | ------ | --------------------------------------------------------------- |
| `prompt`  | string | Text description of the image to generate (max 1000 characters) |

### Optional Parameters

| Parameter             | Type   | Default           | Description                             |
| --------------------- | ------ | ----------------- | --------------------------------------- |
| `num_images`          | number | 1                 | Number of images to generate (1-4)      |
| `image_size`          | string | "landscape\_4\_3" | Image dimensions                        |
| `guidance_scale`      | number | 3.5               | How closely to follow the prompt (1-20) |
| `num_inference_steps` | number | 28                | Number of denoising steps (1-50)        |
| `seed`                | number | random            | Random seed for reproducible results    |

### Image Sizes

| Size             | Dimensions | Aspect Ratio |
| ---------------- | ---------- | ------------ |
| `square_hd`      | 1024×1024  | 1:1          |
| `square`         | 512×512    | 1:1          |
| `portrait_4_3`   | 768×1024   | 3:4          |
| `portrait_16_9`  | 576×1024   | 9:16         |
| `landscape_4_3`  | 1024×768   | 4:3          |
| `landscape_16_9` | 1024×576   | 16:9         |

## Examples

### Basic Image Generation

<CodeGroup>
  ```javascript title="JavaScript" theme={null}
  const response = await fetch('https://firemoon.studio/api/v1/flux/dev', {
    method: 'POST',
    headers: {
      'Authorization': 'Bearer your_api_key',
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      prompt: 'A serene mountain landscape at sunset',
      num_images: 1,
      image_size: 'landscape_16_9'
    })
  });

  const result = await response.json();
  console.log('Image URL:', result.images[0].url);
  ```

  ```python title="Python" theme={null}
  import requests

  response = requests.post(
      'https://firemoon.studio/api/v1/flux/dev',
      headers={
          'Authorization': 'Bearer your_api_key',
          'Content-Type': 'application/json'
      },
      json={
          'prompt': 'A serene mountain landscape at sunset',
          'num_images': 1,
          'image_size': 'landscape_16_9'
      }
  )

  result = response.json()
  print('Image URL:', result['images'][0]['url'])
  ```
</CodeGroup>

### Advanced Generation with Control

<CodeGroup>
  ```javascript title="JavaScript" theme={null}
  const response = await fetch('https://firemoon.studio/api/v1/flux/dev', {
    method: 'POST',
    headers: {
      'Authorization': 'Bearer your_api_key',
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      prompt: 'A futuristic city with flying cars, cyberpunk style, highly detailed',
      num_images: 2,
      image_size: 'landscape_16_9',
      guidance_scale: 7.5, // More creative control
      num_inference_steps: 35, // Higher quality
      seed: 12345 // Reproducible results
    })
  });

  const result = await response.json();
  result.images.forEach((image, index) => {
    console.log(`Image ${index + 1}:`, image.url);
  });
  ```

  ```python theme={null}
  import requests

  def generate_flux_image(prompt, **kwargs):
      response = requests.post(
          'https://firemoon.studio/api/v1/flux/dev',
          headers={
              'Authorization': 'Bearer your_api_key',
              'Content-Type': 'application/json'
          },
          json={
              'prompt': prompt,
              'num_images': kwargs.get('num_images', 1),
              'image_size': kwargs.get('image_size', 'landscape_4_3'),
              'guidance_scale': kwargs.get('guidance_scale', 3.5),
              **kwargs
          }
      )

      response.raise_for_status()
      return response.json()

  # Usage
  result = generate_flux_image(
      'A futuristic city with flying cars, cyberpunk style, highly detailed',
      num_images=2,
      image_size='landscape_16_9',
      guidance_scale=7.5,
      num_inference_steps=35,
      seed=12345
  )

  for image in result['images']:
      print(f"Generated image: {image['url']}")
  ```
</CodeGroup>

## Best Practices

### Prompt Engineering

**Do:**

* Be specific and descriptive
* Include style references (e.g., "in the style of impressionist painting")
* Mention lighting and mood
* Specify composition elements

**Don't:**

* Use negative prompts (not supported)
* Be too vague or abstract
* Include copyrighted character names

### Performance Optimization

1. **Use lower guidance\_scale** for faster generation
2. **Reduce num\_inference\_steps** for speed over quality
3. **Generate multiple images** in one request instead of separate calls
4. **Cache successful seeds** for reproducible variations

### Quality vs Speed Trade-offs

| Use Case        | guidance\_scale | num\_inference\_steps | Expected Time |
| --------------- | --------------- | --------------------- | ------------- |
| Fast draft      | 2.0-3.0         | 20-25                 | 0.5-1.0s      |
| Balanced        | 3.5-5.0         | 25-30                 | 1.0-1.5s      |
| High quality    | 5.0-7.5         | 30-40                 | 1.5-2.5s      |
| Maximum quality | 7.5+            | 40-50                 | 2.5-4.0s      |

## Response Format

```json theme={null}
{
  "images": [
    {
      "url": "https://blob.vercel-storage.com/...",
      "width": 1024,
      "height": 768,
      "content_type": "image/jpeg"
    }
  ],
  "seed": 123456789,
  "has_nsfw_concepts": [false],
  "timings": {
    "inference": 1.2
  }
}
```

## Error Handling

FLUX is generally very reliable, but you may encounter:

* **400 Bad Request**: Invalid parameters or prompt too long
* **429 Too Many Requests**: Rate limit exceeded
* **500 Internal Server Error**: Rare inference failures

Implement proper retry logic for 500 errors and backoff for 429 errors.

## Pricing

FLUX images are billed at our standard image generation rate. Contact us for volume pricing.

<Note>
  FLUX/dev is our recommended model for most use cases due to its excellent balance of speed, quality, and cost.
</Note>