Skip to main content

FLUX.2 Image Generation

FLUX.2 models provide high-quality image generation capabilities through the BFL AI API. This module includes adapters for both FLUX.2 [PRO] and FLUX.2 [FLEX] models, supporting text-to-image generation, image editing, and multi-image composition.

Overview​

  • FLUX.2 [PRO]: High-quality image generation with standard controls
  • FLUX.2 [FLEX]: Flexible generation with advanced controls (guidance scale, steps, prompt upsampling)

Both models support:

  • Text-to-image generation
  • Image editing (prompt + input image)
  • Multi-image composition (up to 8 reference images)
  • Custom width/height control
  • Seed for reproducibility
  • Safety tolerance control

Prerequisites​

  1. BFL AI Account: Sign up at BFL AI
  2. API Key: Obtain your API key from the BFL AI dashboard
  3. Environment Variable: Set BFL_API_KEY in your .env file
BFL_API_KEY=your_bfl_api_key

Installation​

The FLUX.2 adapters are included in the tryon.api module. No additional installation is required beyond the base dependencies.

Quick Start​

FLUX.2 [PRO] - Text-to-Image​

from tryon.api import Flux2ProAdapter

# Initialize adapter
adapter = Flux2ProAdapter()

# Generate image from text prompt
images = adapter.generate_text_to_image(
    prompt="A professional fashion model wearing elegant evening wear on a runway",
    width=1024,
    height=1024,
    seed=42
)

# Save result
images[0].save("outputs/flux2_pro_result.png")

FLUX.2 [FLEX] - Advanced Generation​

from tryon.api import Flux2FlexAdapter

# Initialize adapter
adapter = Flux2FlexAdapter()

# Generate with advanced controls
images = adapter.generate_text_to_image(
    prompt="A stylish fashion model wearing elegant evening wear",
    width=1024,
    height=1024,
    guidance=7.5,  # Higher guidance = more adherence to prompt
    steps=50,  # More steps = higher quality
    prompt_upsampling=True,
    seed=42
)

images[0].save("outputs/flux2_flex_result.png")

API Reference​

Flux2ProAdapter​

__init__(api_key=None, base_url=None)​

Initialize the FLUX.2 [PRO] adapter.

Parameters:

  • api_key (str, optional): BFL API key. Defaults to BFL_API_KEY environment variable.
  • base_url (str, optional): Base URL for BFL API. Defaults to https://api.bfl.ai.

Raises:

  • ValueError: If API key is not provided.

Example:

# Using environment variable
adapter = Flux2ProAdapter()

# Using parameter
adapter = Flux2ProAdapter(api_key="your_api_key")

generate_text_to_image(prompt, width=None, height=None, seed=None, safety_tolerance=2, output_format="png", **kwargs)​

Generate image from text prompt.

Parameters:

  • prompt (str, required): Text description of the image to generate
  • width (int, optional): Width of output image (minimum: 64)
  • height (int, optional): Height of output image (minimum: 64)
  • seed (int, optional): Seed for reproducibility
  • safety_tolerance (int): Moderation tolerance (0-5). 0 = most strict, 5 = least strict. Default: 2
  • output_format (str): Output format. Options: "jpeg", "png". Default: "png"
  • **kwargs: Additional parameters (webhook_url, webhook_secret, etc.)

Returns:

  • List[Image.Image]: List of PIL Image objects

Example:

images = adapter.generate_text_to_image(
    prompt="A professional fashion model wearing elegant evening wear",
    width=1024,
    height=1024,
    seed=42
)

generate_image_edit(prompt, input_image, width=None, height=None, seed=None, safety_tolerance=2, output_format="png", **kwargs)​

Generate edited image from prompt and input image.

Parameters:

  • prompt (str, required): Text description of how to edit the image
  • input_image (Union[str, io.BytesIO, Image.Image], required): Input image (file path, URL, PIL Image, file-like object, or base64)
  • width (int, optional): Width of output image (minimum: 64)
  • height (int, optional): Height of output image (minimum: 64)
  • seed (int, optional): Seed for reproducibility
  • safety_tolerance (int): Moderation tolerance (0-5). Default: 2
  • output_format (str): Output format. Options: "jpeg", "png". Default: "png"
  • **kwargs: Additional parameters

Returns:

  • List[Image.Image]: List of PIL Image objects

Example:

images = adapter.generate_image_edit(
    prompt="Change the outfit to casual streetwear style",
    input_image="model.jpg",
    width=1024,
    height=1024
)

generate_multi_image(prompt, images, width=None, height=None, seed=None, safety_tolerance=2, output_format="png", **kwargs)​

Generate image from multiple input images (composition & style transfer).

Parameters:

  • prompt (str, required): Text description of how to combine/transform the images
  • images (List[Union[str, io.BytesIO, Image.Image]], required): List of input images (file paths, URLs, PIL Images, file-like objects, or base64). Maximum 8 images supported.
  • width (int, optional): Width of output image (minimum: 64)
  • height (int, optional): Height of output image (minimum: 64)
  • seed (int, optional): Seed for reproducibility
  • safety_tolerance (int): Moderation tolerance (0-5). Default: 2
  • output_format (str): Output format. Options: "jpeg", "png". Default: "png"
  • **kwargs: Additional parameters

Returns:

  • List[Image.Image]: List of PIL Image objects

Raises:

  • ValueError: If more than 8 images are provided.

Example:

images = adapter.generate_multi_image(
    prompt="Create a fashion catalog layout combining these clothing styles",
    images=["outfit1.jpg", "outfit2.jpg", "accessories.jpg"],
    width=1024,
    height=1024
)

Flux2FlexAdapter​

__init__(api_key=None, base_url=None)​

Initialize the FLUX.2 [FLEX] adapter.

Parameters:

  • api_key (str, optional): BFL API key. Defaults to BFL_API_KEY environment variable.
  • base_url (str, optional): Base URL for BFL API. Defaults to https://api.bfl.ai.

Raises:

  • ValueError: If API key is not provided.

generate_text_to_image(prompt, width=None, height=None, seed=None, guidance=3.5, steps=28, prompt_upsampling=True, safety_tolerance=2, output_format="png", **kwargs)​

Generate image from text prompt with advanced controls.

Parameters:

  • prompt (str, required): Text description of the image to generate
  • width (int, optional): Width of output image (minimum: 64)
  • height (int, optional): Height of output image (minimum: 64)
  • seed (int, optional): Seed for reproducibility
  • guidance (float): Guidance scale (1.5-10). Higher values = more adherence to prompt. Default: 3.5
  • steps (int): Number of generation steps. More steps = higher quality. Default: 28
  • prompt_upsampling (bool): Whether to use prompt upsampling. Default: True
  • safety_tolerance (int): Moderation tolerance (0-5). Default: 2
  • output_format (str): Output format. Options: "jpeg", "png". Default: "png"
  • **kwargs: Additional parameters (webhook_url, webhook_secret, input_image_blob_path, etc.)

Returns:

  • List[Image.Image]: List of PIL Image objects

Raises:

  • ValueError: If guidance is not between 1.5 and 10.

Example:

images = adapter.generate_text_to_image(
    prompt="A professional fashion model wearing elegant evening wear",
    width=1024,
    height=1024,
    guidance=7.5,
    steps=50,
    prompt_upsampling=True,
    seed=42
)

generate_image_edit(prompt, input_image, width=None, height=None, seed=None, guidance=3.5, steps=28, prompt_upsampling=True, safety_tolerance=2, output_format="png", **kwargs)​

Generate edited image from prompt and input image with advanced controls.

Parameters:

  • prompt (str, required): Text description of how to edit the image
  • input_image (Union[str, io.BytesIO, Image.Image], required): Input image
  • width (int, optional): Width of output image (minimum: 64)
  • height (int, optional): Height of output image (minimum: 64)
  • seed (int, optional): Seed for reproducibility
  • guidance (float): Guidance scale (1.5-10). Default: 3.5
  • steps (int): Number of generation steps. Default: 28
  • prompt_upsampling (bool): Whether to use prompt upsampling. Default: True
  • safety_tolerance (int): Moderation tolerance (0-5). Default: 2
  • output_format (str): Output format. Options: "jpeg", "png". Default: "png"
  • **kwargs: Additional parameters

Returns:

  • List[Image.Image]: List of PIL Image objects

generate_multi_image(prompt, images, width=None, height=None, seed=None, guidance=3.5, steps=28, prompt_upsampling=True, safety_tolerance=2, output_format="png", **kwargs)​

Generate image from multiple input images with advanced controls.

Parameters:

  • prompt (str, required): Text description of how to combine/transform the images
  • images (List[Union[str, io.BytesIO, Image.Image]], required): List of input images (maximum 8)
  • width (int, optional): Width of output image (minimum: 64)
  • height (int, optional): Height of output image (minimum: 64)
  • seed (int, optional): Seed for reproducibility
  • guidance (float): Guidance scale (1.5-10). Default: 3.5
  • steps (int): Number of generation steps. Default: 28
  • prompt_upsampling (bool): Whether to use prompt upsampling. Default: True
  • safety_tolerance (int): Moderation tolerance (0-5). Default: 2
  • output_format (str): Output format. Options: "jpeg", "png". Default: "png"
  • **kwargs: Additional parameters

Returns:

  • List[Image.Image]: List of PIL Image objects

Supported Input Formats​

All image input parameters accept:

  • File paths (str): Path to local image file
  • URLs (str): HTTP/HTTPS URL to image
  • Base64 strings (str): Base64-encoded image data
  • File-like objects (io.BytesIO): BytesIO or similar object
  • PIL Images (Image.Image): PIL Image object

Async Processing​

FLUX.2 API processes requests asynchronously. The adapters automatically:

  1. Submit the request and receive a task_id and polling_url
  2. Poll the task status endpoint until completion
  3. Return PIL Image objects when the task succeeds
  4. Raise errors if the task fails or times out (default timeout: 5 minutes)

Examples​

Complete Workflow​

from tryon.api import Flux2ProAdapter, Flux2FlexAdapter

# FLUX.2 [PRO] - Simple text-to-image
pro_adapter = Flux2ProAdapter()
pro_images = pro_adapter.generate_text_to_image(
    prompt="A professional fashion model wearing elegant evening wear",
    width=1024,
    height=1024
)
pro_images[0].save("outputs/pro_result.png")

# FLUX.2 [FLEX] - Advanced generation
flex_adapter = Flux2FlexAdapter()
flex_images = flex_adapter.generate_text_to_image(
    prompt="A stylish fashion model wearing elegant evening wear",
    width=1024,
    height=1024,
    guidance=8.0,
    steps=50,
    prompt_upsampling=True
)
flex_images[0].save("outputs/flex_result.png")

# Multi-image composition
composition_images = pro_adapter.generate_multi_image(
    prompt="Combine these clothing items into a cohesive fashion outfit",
    images=["top.jpg", "pants.jpg", "jacket.jpg", "shoes.jpg"],
    width=1024,
    height=1024
)
composition_images[0].save("outputs/composition_result.png")

Image Editing​

from tryon.api import Flux2FlexAdapter

adapter = Flux2FlexAdapter()

# Edit image with advanced controls
images = adapter.generate_image_edit(
    prompt="Transform the outfit to match a vintage 1920s fashion style",
    input_image="model.jpg",
    guidance=8.0,
    steps=50,
    prompt_upsampling=True,
    seed=42
)

images[0].save("outputs/edited_result.png")

Error Handling​

The adapters raise ValueError for:

  • Missing API key
  • Invalid parameters (e.g., guidance out of range)
  • API errors (HTTP errors, task failures)
  • Timeout errors (task takes too long)
  • Invalid image formats

Example:

try:
    images = adapter.generate_text_to_image(
        prompt="A fashion model",
        guidance=15  # Invalid: must be between 1.5 and 10
    )
except ValueError as e:
    print(f"Error: {e}")

Best Practices​

  1. Prompt Quality: Write detailed, specific prompts for better results
  2. Guidance Scale: Start with default (3.5) and adjust based on results
    • Lower (1.5-3): More creative, less strict to prompt
    • Higher (7-10): More strict adherence to prompt
  3. Steps: More steps = higher quality but slower generation
    • Default (28): Good balance
    • High quality (50+): Slower but better results
  4. Seed: Use seeds for reproducible results
  5. Safety Tolerance: Adjust based on your content moderation needs
  6. Multi-Image: Use up to 8 reference images for best composition results

References​