API reference for Flux Krea. FLUX.1 Krea [dev] is a 12 billion parameter flow transformer that generates high-quality images from text with incredible aesthetics. It is suitable for personal and comme
import fal_clientdef on_queue_update(update): if isinstance(update, fal_client.InProgress): for log in update.logs: print(log["message"])result = fal_client.subscribe( "fal-ai/flux/krea", arguments={ "prompt": "A candid street photo of a woman with a pink bob and bold eyeliner on a graffiti-covered subway platform. She wears a bright yellow patent leather coat over a black-and-white checkered turtleneck and platform boots. Natural subway lighting creates an authentic urban scene with a relaxed, unposed feel." }, with_logs=True, on_queue_update=on_queue_update,)print(result)
The CFG (Classifier Free Guidance) scale is a measure of how close you want
the model to stick to your prompt when looking for a related image to show you. Default value: 4.5Range: 1 to 20
{ "prompt": "A candid street photo of a woman with a pink bob and bold eyeliner on a graffiti-covered subway platform. She wears a bright yellow patent leather coat over a black-and-white checkered turtleneck and platform boots. Natural subway lighting creates an authentic urban scene with a relaxed, unposed feel.", "image_size": "landscape_4_3", "num_inference_steps": 28, "guidance_scale": 4.5, "sync_mode": false, "num_images": 1, "enable_safety_checker": true, "output_format": "jpeg", "acceleration": "none"}
import fal_clientdef on_queue_update(update): if isinstance(update, fal_client.InProgress): for log in update.logs: print(log["message"])result = fal_client.subscribe( "fal-ai/flux/krea/image-to-image", arguments={ "image_url": "https://fal.media/files/koala/Chls9L2ZnvuipUTEwlnJC.png", "prompt": "A cat dressed as a wizard with a background of a mystic forest." }, with_logs=True, on_queue_update=on_queue_update,)print(result)
The CFG (Classifier Free Guidance) scale is a measure of how close you want
the model to stick to your prompt when looking for a related image to show you. Default value: 4.5Range: 1 to 20
{ "image_url": "https://fal.media/files/koala/Chls9L2ZnvuipUTEwlnJC.png", "strength": 0.95, "num_inference_steps": 40, "prompt": "A cat dressed as a wizard with a background of a mystic forest.", "guidance_scale": 4.5, "sync_mode": false, "num_images": 1, "enable_safety_checker": true, "output_format": "jpeg", "acceleration": "none"}
The CFG (Classifier Free Guidance) scale is a measure of how close you want
the model to stick to your prompt when looking for a related image to show you. Default value: 4.5Range: 1 to 20
Black Forest Labs’ FLUX.1 Krea [dev] delivers high-quality image generation from a 12 billion parameter flow transformer at $0.025 per megapixel. Trading raw parameter count for aesthetic refinement through collaboration with Krea AI, this model prioritizes visual quality and commercial viability over pure speed. Built for creators and enterprises needing production-ready images without the licensing restrictions of research-only alternatives.Use Cases: Commercial Product Photography | Marketing Asset Generation | Creative Concept Visualization
FLUX.1 Krea [dev] is a commercially-licensed alternative in the FLUX family, offering flexible resolution scaling with per-megapixel pricing that becomes cost-competitive at lower resolutions, delivering 40 generations per $1.00 at standard sizes.
Metric
Result
Context
Parameter Count
12 billion
Flow transformer architecture optimized for aesthetic quality
Inference Steps
28 (default)
Configurable 1-50 steps for speed/quality tradeoff
Cost per Megapixel
$0.025
40 generations per $1.00 at 1MP; images billed by rounding up to nearest megapixel
Max Resolution
Up to 4 megapixels
Flexible aspect ratios via landscape/portrait/square presets
FLUX.1 Krea [dev] implements a 12 billion parameter flow transformer architecture tuned specifically for aesthetic quality through Black Forest Labs’ collaboration with Krea AI. Unlike research-licensed models that restrict commercial use, this variant ships with full commercial permissions while maintaining the prompt interpretation capabilities of the base FLUX.1 architecture.What this means for you:
Flexible Resolution Scaling: Generate from sub-megapixel thumbnails to 4MP hero images with per-megapixel billing, paying only for the resolution you need rather than flat per-image costs
Configurable Quality Control: Adjust inference steps (1-50) and guidance scale (1-20) to balance generation speed against prompt adherence for different production workflows
Built-in Safety Filtering: Optional safety checker validates outputs against NSFW concepts before delivery, reducing moderation overhead for customer-facing applications
Batch Efficiency: Generate up to 4 variations per API call with consistent seed control for A/B testing creative concepts without multiple requests
FLUX.1 [dev] ($0.025/MP) – FLUX.1 Krea [dev] trades the base model’s research-only license for commercial permissions while adding aesthetic tuning from Krea AI’s dataset at identical pricing. FLUX.1 [dev] remains ideal for research applications where licensing restrictions aren’t a concern.FLUX.1 SRPO [dev] ($0.025/MP) – FLUX.1 Krea [dev] prioritizes aesthetic quality through Krea’s training data, while SRPO emphasizes style refinement through reinforcement learning. Both share commercial licensing and per-megapixel pricing, with SRPO offering stronger stylistic control for brand-consistent generation.**Flux 2 Pro (0.055/MP)∗∗–FLUX.1Krea[dev]deliverscommercially−licensedgenerationat2.2xlowercost(0.025 vs $0.055 per megapixel). Flux 2 Pro provides next-generation prompt understanding and detail preservation for applications where maximum quality justifies the premium pricing.**Flux 2 Flex (0.035/MP)∗∗–FLUX.1Krea[dev]offers1.4xcostsavings(0.025 vs $0.035 per megapixel) while maintaining commercial licensing. Flux 2 Flex balances the quality improvements of Flux 2 architecture with more accessible pricing for high-volume generation workflows.
Documentation
Examples
Model API Reference
SDK Reference
Platform API
Changelog