Black Forest Labs

Access Black Forest Labs' FLUX models including Pro, Ultra, Tools, and Kontext variants for advanced image generation through Runware's unified API. Learn about BFL-specific parameters, limitations, and technical specifications.

Introduction

Black Forest Labs' AI models are integrated into the Runware platform through our unified API, providing access to advanced image generation technology through the professional FLUX model family.

Black Forest Labs, founded by the creators of Stable Diffusion, offers commercial-grade FLUX models that deliver superior prompt adherence, visual quality, and specialized capabilities for professional workflows. Through the providerSettings.bfl object, you can access unique features such as prompt upsampling, raw output modes, and adjustable safety tolerance, while maintaining the consistency of Runware's standard API structure. This page documents the technical specifications, parameter requirements, and provider-specific settings for all Black Forest Labs models available through our platform.

providerSettings
object

Contains provider-specific configuration settings that customize the behavior of different AI models and services. Each provider has its own set of parameters that control various aspects of the generation process.

The providerSettings parameter is an object that contains nested objects for each supported provider.

View example 
{
  "taskType": "imageInference",
  "taskUUID": "991e641a-d2a8-4aa3-9883-9d6fe230fff8",
  "positivePrompt": "a beautiful landscape with mountains",
  "model": "bfl:2@2",
  "providerSettings": { 
    "bfl": { 
      "promptUpsampling": true,
      "safetyTolerance": 4,
      "raw": false
    } 
  } 
}
Properties ⁨1⁩ property
providerSettings » bfl
bfl
object

Configuration object for Black Forest Labs (BFL) specific features. BFL models offer advanced prompt processing and content safety controls.

View example 
{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "positivePrompt": "a beautiful landscape at sunset",
  "model": "bfl:1@1",
  "width": 1024,
  "height": 1024,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true,
      "safetyTolerance": 6
    }
  }
}
Properties ⁨3⁩ properties
providerSettings » bfl » promptUpsampling
promptUpsampling
boolean Default: false

Enables automatic enhancement and expansion of the input prompt to improve generation quality and detail.

When enabled, BFL's prompt upsampling system analyzes your text description and adds relevant details and descriptive elements that enhance the final output without changing the core intent of your prompt.

providerSettings » bfl » safetyTolerance
safetyTolerance
integer Min: 0 Max: 6 Default: 2

Controls the tolerance level for input and output content moderation. Lower values apply stricter content filtering, while higher values are more permissive. Range from 0 (most strict) to 6 (least strict).

providerSettings » bfl » raw
raw
boolean Default: false

Controls the level of post-processing applied to generated images.

When enabled, the raw mode produces images that are closer to the model's direct output without additional processing layers. This can result in more natural-looking images but may sacrifice some visual polish and consistency that post-processing typically provides.

Image models

FLUX.1.1 Pro

Black Forest Labs' FLUX.1.1 Pro model delivers enhanced image quality, improved prompt adherence, and increased output diversity with faster and more accurate results.

Model AIR ID: bfl:2@1.

Supported workflows: Text-to-image.

Technical specifications:

  • Positive prompt: 2-3000 characters.
  • Supported dimensions: 256-1440 pixels (width and height).

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d1",
  "model": "bfl:2@1",
  "positivePrompt": "Architectural visualization of sustainable building design with green technology integration",
  "width": 1280,
  "height": 720
}

FLUX.1.1 Pro Ultra

Black Forest Labs' FLUX.1.1 Pro Ultra model provides high-resolution generation up to 4 megapixels with raw mode for detailed and realistic outputs.

Model AIR ID: bfl:2@2.

Supported workflows: Text-to-image.

Technical specifications:

  • Positive prompt: 2-3000 characters.
  • Supported dimensions: 3136×1344 (21:9), 2752×1536 (16:9), 2368×1792 (4:3), 2496×1664 (3:2), 2048×2048 (1:1), 1664×2496 (2:3), 1792×2368 (3:4), 1536×2752 (9:16), 1344×3136 (9:21).

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance, raw.

The raw setting generates less processed, more natural-looking images when enabled.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d1",
  "model": "bfl:2@2",
  "positivePrompt": "Ultra-detailed landscape photography of mountain vista during golden hour",
  "width": 2752,
  "height": 1536,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true,
      "raw": true
    }
  }
}

FLUX.1 Fill Pro

Black Forest Labs' FLUX.1 Fill Pro model provides advanced inpainting and outpainting capabilities, enabling seamless editing and expansion of images based on text prompts.

Model AIR ID: bfl:1@2.

Supported workflows: Inpainting.

Technical specifications:

  • Positive prompt: 2-3000 characters.
  • Output dimensions: Automatically determined based on input image.
  • CFG Scale: 1.5-100 (default: 60).
  • Steps: 15-50 (default: 50).
  • Seed image: Required.
  • Mask image: Required.

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d1",
  "model": "bfl:1@2",
  "positivePrompt": "Modern furniture in contemporary living room setting",
  "seedImage": "c64351d5-4c59-42f7-95e1-eace013eddab",
  "maskImage": "d7e8f9a0-2b5c-4e7f-a1d3-9c8b7a6e5d4f",
  "steps": 40,
  "CFGScale": 60
}

FLUX.1 Expand Pro

Black Forest Labs' FLUX.1 Expand Pro model specializes in outpainting, designed to extend images beyond their original borders while maintaining visual coherence and style.

Model AIR ID: bfl:1@3.

Supported workflows: Outpainting.

Technical specifications:

  • Positive prompt: 2-3000 characters.
  • Output dimensions: Automatically determined based on input image and expansion.
  • CFG Scale: 1.5-100 (default: 60).
  • Steps: 15-50 (default: 50).
  • Seed image: Required.
  • Outpaint directions: Required (maximum 2048 pixels per side).

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d1",
  "model": "bfl:1@3",
  "positivePrompt": "Extended natural landscape with mountains and forest",
  "seedImage": "c64351d5-4c59-42f7-95e1-eace013eddab",
  "outpaint": {
    "top": 256,
    "bottom": 256,
    "left": 512,
    "right": 512
  },
  "steps": 45,
  "CFGScale": 45,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true
    }
  }
}

FLUX.1 Kontext [pro]

Black Forest Labs' FLUX.1 Kontext [pro] model enables fast, iterative editing with local and full-scene changes, preserving style across multiple modifications.

Model AIR ID: bfl:3@1.

Supported workflows: Text-to-image, reference-to-image.

Technical specifications:

  • Positive prompt: 2-3000 characters.
  • Supported dimensions: 1568×672 (21:9), 1392×752 (16:9), 1184×880 (4:3), 1248×832 (3:2), 1024×1024 (1:1), 832×1248 (2:3), 880×1184 (3:4), 752×1392 (9:16), 672×1568 (9:21).
  • Reference images: Supports referenceImages with 2 images.

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d1",
  "model": "bfl:3@1",
  "positivePrompt": "Transform the scene to include autumn foliage and warm lighting",
  "width": 1392,
  "height": 752
}

FLUX.1 Kontext [max]

Black Forest Labs' FLUX.1 Kontext [max] model delivers the highest quality and prompt accuracy with faster, sharper edits and premium typography support.

Model AIR ID: bfl:4@1.

Supported workflows: Text-to-image, reference-to-image.

Technical specifications:

  • Positive prompt: 2-3000 characters.
  • Supported dimensions: 1568×672 (21:9), 1392×752 (16:9), 1184×880 (4:3), 1248×832 (3:2), 1024×1024 (1:1), 832×1248 (2:3), 880×1184 (3:4), 752×1392 (9:16), 672×1568 (9:21).
  • Reference images: Supports referenceImages with 2 images.

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d1",
  "model": "bfl:4@1",
  "positivePrompt": "Professional logo design with premium typography and sophisticated branding elements",
  "width": 1024,
  "height": 1024,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true
    }
  }
}

FLUX.2 [dev]

Black Forest Labs' FLUX.2 [dev] model is the open weights release offering full architectural control for developers who need to experiment with custom pipelines and advanced workflows. Available through Runware with optimized inference for fast, cost-effective generation while maintaining complete control over sampling behavior and guidance strategies.

Model AIR ID: runware:400@1.

Supported workflows: Text-to-image, reference-to-image.

Technical specifications:

  • Positive prompt: 1-10000 characters.
  • Supported dimensions: 512-2048 pixels (width and height, multiples of 16).
  • Reference images: Supports up to 4 images via referenceImages.
  • CFG Scale: 1-20 (default: 4).
  • Steps: 1-50.
  • Acceleration: Supports none, low, medium or high values (Default: medium).
{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d2",
  "model": "runware:400@1",
  "positivePrompt": "Experimental digital art with abstract geometric patterns and vibrant color gradients",
  "width": 1024,
  "height": 1024,
  "steps": 30,
  "CFGScale": 5
}
{
  "taskType": "imageInference",
  "taskUUID": "6ba7b829-9dad-11d1-80b4-00c04fd430c8",
  "model": "runware:400@1",
  "positivePrompt": "Detailed fantasy landscape with magical elements and atmospheric lighting",
  "width": 1536,
  "height": 864,
  "steps": 20,
  "CFGScale": 5,
  "acceleration": "high"
}
{
  "taskType": "imageInference",
  "taskUUID": "550e8400-e29b-41d4-a716-446655440010",
  "model": "runware:400@1",
  "positivePrompt": "Character in similar style and composition to reference images",
  "inputs": {
    "referenceImages": [
      "c64351d5-4c59-42f7-95e1-eace013eddab",
      "d7e8f9a0-2b5c-4e7f-a1d3-9c8b7a6e5d4f"
    ]
  },
  "width": 768,
  "height": 1024,
  "steps": 35,
  "CFGScale": 4.0
}

FLUX.2 [pro]

Black Forest Labs' FLUX.2 [pro] model delivers high-quality production-ready output with robust reference-image editing capabilities. Built for reliability and speed, this model excels at consistent results across high-volume workflows where dependable performance matters more than deep customization.

Model AIR ID: bfl:5@1.

Supported workflows: Text-to-image, reference-to-image.

Technical specifications:

  • Positive prompt: 1-3000 characters.
  • Supported dimensions: 256-1920 pixels (width and height, multiples of 16).
  • Reference images: Supports up to 9 images via referenceImages.
  • Total input capacity: 9 megapixels.

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d3",
  "model": "bfl:5@1",
  "positivePrompt": "Professional product photography of modern electronics on clean white background",
  "width": 1280,
  "height": 720,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true
    }
  }
}
{
  "taskType": "imageInference",
  "taskUUID": "6ba7b830-9dad-11d1-80b4-00c04fd430c8",
  "model": "bfl:5@1",
  "positivePrompt": "Maintain consistent lighting and brand aesthetic across product variations",
  "inputs": {
    "referenceImages": [
      "c64351d5-4c59-42f7-95e1-eace013eddab",
      "d7e8f9a0-2b5c-4e7f-a1d3-9c8b7a6e5d4f",
      "e8f9a0b1-3c6d-4e8f-b2e4-0d9e8f7c6b5a"
    ]
  },
  "width": 1024,
  "height": 1024
}
{
  "taskType": "imageInference",
  "taskUUID": "550e8400-e29b-41d4-a716-446655440011",
  "model": "bfl:5@1",
  "positivePrompt": "Cinematic scene with detailed environment and professional composition",
  "width": 1920,
  "height": 1080,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true,
      "safetyTolerance": 2
    }
  }
}
FLUX.2 [pro] Pricing

Base rates:

  • Input (reference images): $0.015 per megapixel.
  • Output: First megapixel is $0.03, then $0.015 per additional megapixel.

How megapixels are calculated: A megapixel (MP) is 1,048,576 pixels (1024×1024). To calculate MP for any image, multiply width × height and divide by 1,048,576. Pricing rounds up each image separately to the nearest whole MP.

Common resolutions:

  • 512×512 = 0.26 MP → rounds to 1 MP.
  • 1280×720 (720p) = 0.92 MP → rounds to 1 MP.
  • 1024×1024 = 1 MP → stays 1 MP.
  • 1920×1080 (1080p) = 2.07 MP → rounds to 3 MP.

Reference image processing:

  • Single reference: Preserves original resolution up to 4 MP, downsamples above 4 MP.
  • Multiple references: Each reference is automatically downscaled to 1 MP if original exceeds 1 MP.
  • Each reference image is billed as 1 MP minimum, regardless of actual resolution.
  • Maximum output resolution: 4 MP.

Example costs:

  • Text-to-image at 1024×1024: 1 MP output = $0.03.
  • Text-to-image at 1920×1080: 3 MP output = $0.03 + (2 × $0.015) = $0.06.
  • Image-to-image with one 1920×1080 reference at 1024×1024 output: 3 MP input + 1 MP output = (3 × $0.015) + $0.03 = $0.075.
  • Image-to-image with three references at 1024×1024 output: 3 MP input (3×1 MP) + 1 MP output = (3 × $0.015) + $0.03 = $0.075.

FLUX.2 [flex]

Black Forest Labs' FLUX.2 [flex] model provides the strongest text rendering accuracy in the FLUX family with fine-grained control over composition and layout. Designed specifically for branded design, posters, and typography-driven work, this model maintains character and product consistency across multiple reference images with precise placement and readability.

Model AIR ID: bfl:6@1.

Supported workflows: Text-to-image, reference-to-image.

Technical specifications:

  • Positive prompt: 1-3000 characters.
  • Supported dimensions: 256-1920 pixels (width and height, multiples of 16).
  • Reference images: Supports up to 10 images via referenceImages.
  • Total input capacity: 14 megapixels.
  • CFG Scale: 1-20 (default: 2.5).
  • Steps: 1-50.

Provider-specific settings:

Parameters supported: promptUpsampling, safetyTolerance.

{
  "taskType": "imageInference",
  "taskUUID": "24cd5dff-cb81-4db5-8506-b72a9425f9d4",
  "model": "bfl:6@1",
  "positivePrompt": "Modern poster with bold typography 'INNOVATION SUMMIT 2025' and structured layout with geometric elements",
  "width": 1024,
  "height": 1440,
  "steps": 28,
  "CFGScale": 6
}
{
  "taskType": "imageInference",
  "taskUUID": "6ba7b831-9dad-11d1-80b4-00c04fd430c8",
  "model": "bfl:6@1",
  "positivePrompt": "Product packaging design maintaining consistent brand character and visual identity",
  "inputs": {
    "referenceImages": [
      "c64351d5-4c59-42f7-95e1-eace013eddab",
      "d7e8f9a0-2b5c-4e7f-a1d3-9c8b7a6e5d4f",
      "e8f9a0b1-3c6d-4e8f-b2e4-0d9e8f7c6b5a",
      "f9a0b1c2-4d7e-5f9g-c3f5-1e0f9g8d7c6b"
    ]
  },
  "width": 768,
  "height": 1024,
  "providerSettings": {
    "bfl": {
      "promptUpsampling": true
    }
  }
}
{
  "taskType": "imageInference",
  "taskUUID": "550e8400-e29b-41d4-a716-446655440012",
  "model": "bfl:6@1",
  "positivePrompt": "Clean infographic layout with precise text placement, data visualization elements, and professional typography",
  "width": 1280,
  "height": 720,
  "steps": 30,
  "CFGScale": 5,
  "providerSettings": {
    "bfl": {
      "safetyTolerance": 2
    }
  }
}
FLUX.2 [flex] Pricing

Base rate: $0.06 per megapixel for both input and output images.

How megapixels are calculated: A megapixel (MP) is 1,048,576 pixels (1024×1024). To calculate MP for any image, multiply width × height and divide by 1,048,576. Pricing rounds up each image separately to the nearest whole MP.

Common resolutions:

  • 512×512 = 0.26 MP → rounds to 1 MP.
  • 1280×720 (720p) = 0.92 MP → rounds to 1 MP.
  • 1024×1024 = 1 MP → stays 1 MP.
  • 1920×1080 (1080p) = 2.07 MP → rounds to 3 MP.

Reference image processing:

  • Single reference: Preserves original resolution up to 4 MP, downsamples above 4 MP.
  • Multiple references: Each reference is automatically downscaled to 1 MP if original exceeds 1 MP.
  • Each reference image is billed as 1 MP minimum, regardless of actual resolution.
  • Maximum output resolution: 4 MP.

Example costs:

  • Text-to-image at 1024×1024: 1 MP output = $0.06.
  • Image-to-image with one 1920×1080 reference: 3 MP input + 3 MP output = $0.36.
  • Image-to-image with three reference images: 3 MP input (3×1 MP) + 1 MP output = $0.24.
Ask AI
Context: Full page
Send feedback
Context: Full page