Five New FLUX Models Live on WaveSpeedAI: From Creative Variations to Precision Control

WaveSpeedAI,Fri Apr 18 2025

This week, we’re introducing five new FLUX model endpoints—each tailored for a specific creative or structural task. Whether you’re looking to generate consistent image variations, apply advanced style transfers, or control generation with edge or depth maps, these models expand your creative toolkit with precision and speed.

Below is a breakdown of what each new endpoint offers, how it works, and potential application scenarios.

FLUX Models

✨ FLUX Redux Dev

Model: wavespeed-ai/flux-redux-dev
Type: Image-to-Image (Open Weight)

FLUX Redux Dev is an open-weight image-to-image variation model designed to preserve the key structure of an input image while generating creative new variants. Built on the FLUX.1 development architecture, this model is ideal for experimentation and deployment in workflows requiring flexible, fast generation.

Key Features:

Open weight model for flexible use and customization
Maintains core elements of the original image
Ideal for fast visual iteration across creative directions

Use Cases:

Design exploration
Content variation for A/B testing
Visual experimentation in branding and illustration

🌀 FLUX Dev Fill

Model: wavespeed-ai/flux-dev-fill
Type: Image-to-Image

FLUX Dev Fill is a high-performance endpoint built on FLUX.1 [dev], optimized for rapid image transformation. It enables detailed style transfers and image modifications while maintaining the underlying image semantics. Dev Fill is ideal for high-speed, high-fidelity workflows during the prototyping phase.

Key Features:

Fast style transfer with structural preservation
Reliable output quality across diverse inputs
Seamlessly integrates with the FLUX development stack

Use Cases:

Visual style testing for media and entertainment
Concept art and digital mockups
High-volume creative workflows

🧩 FLUX Pro Redux

Model: wavespeed-ai/flux-pro-redux
Type: Image-to-Image

FLUX Pro Redux is the professional-grade endpoint of the FLUX.1 [pro] architecture. It provides enhanced fidelity and control for image-to-image transformations, preserving semantic meaning and visual consistency. It’s designed for use in demanding environments where output quality must meet production standards.

Key Features:

High-resolution generation with stable structure
Precision in visual consistency and style
Tailored for professional creative pipelines

Use Cases:

Product design visualization
Marketing asset creation
Concept development for film, fashion, and media

🖋️ FLUX Control LoRA Canny

Model: wavespeed-ai/flux-control-lora-canny
Type: Image-to-Image (Control Image Input: Canny Edges)

FLUX Control LoRA Canny enables precise structure control via Canny edge maps. This endpoint interprets edge input as a guide for composition, allowing users to direct the generated content using structural sketches. It’s particularly useful for workflows that begin with shape or silhouette design.

Key Features:

Uses Canny edge maps for structural guidance
Supports detailed generation with accurate outline adherence
Integrates with LoRA for enhanced control

Use Cases:

Sketch-to-image generation
Industrial design concepts
Visual development in animation and comics

🌐 FLUX Control LoRA Depth

Model: wavespeed-ai/flux-control-lora-depth
Type: Image-to-Image (Control Image Input: Depth Map)

FLUX Control LoRA Depth uses depth maps to guide image generation, allowing for realistic perspective, spatial relationships, and 3D-aware composition. The model is optimized for scenarios where depth cues significantly influence the final output.

Key Features:

Depth-guided image generation
Preserves spatial coherence and camera perspective
Robust output with layered composition

Use Cases:

Scene layout for visual storytelling
3D-to-2D conversion workflows
Architectural visualization and interior concept design

Each of these endpoints enhances the modularity of the FLUX stack and offers new tools for creative and technical professionals alike. You can explore and test them directly via WaveSpeedAI model store or integrate them into your own pipelines using our API.