FLUX.1 Redux

Replicate
Image Generation

FLUX image-variation adapter. Generate variations and remixes from a reference image.

Generate with FLUX.1 Redux
Describe what you want and pick a size β€” the image renders inline.
Sign in to try this model with €5 free credits.
Sign in
Result appears here
TL;DRΒ·Last updated May 16, 2026

FLUX.1 Redux is image generation AI model from Replicate, priced at €0.000 per 1M input tokens with a unknown context window.

Try FLUX.1 Redux
Sign in to generate β€” 50 free credits on sign-up

Pricing

Price per Generation
Per generation€0.03

API Integration

Use our OpenAI-compatible API to integrate FLUX.1 Redux into your application.

Install
npm install railwail
JavaScript / TypeScript
import railwail from "railwail";

const rw = railwail("YOUR_API_KEY");

const images = await rw.run("flux-1-redux", "A beautiful sunset over Tokyo");
console.log(images[0].url);

// Or use the image() method for full control
const res = await rw.image("flux-1-redux", "A cat in space", {
  size: "1024x1024",
  n: 1,
});
console.log(res.data[0].url);
Specifications
Price
€0.03
Developer
Replicate
Category
Image Generation
Supported Formats
image
text
Tags
flux
black-forest-labs
image-edit
variation
image-to-image
remix

Deep dive β€” Black Forest Labs's FLUX.1 Redux

About Black Forest Labs
Founded 2024 Β· Freiburg, Germany

Black Forest Labs was founded in August 2024 in Freiburg, Germany by Robin Rombach, Patrick Esser, Andreas Blattmann and Dominik Lorenz, the original Stable Diffusion team. After raising a $31M seed round led by Andreessen Horowitz, the lab released the FLUX.1 family (pro, dev, schnell) and the FLUX.1 Tools (Fill, Canny, Depth, Redux). Their mission is to be the leading European open foundation-model lab for generative media.

Visit Black Forest Labs β†’
Architecture
Rectified-flow DiT with image-embedding (Redux) conditioning for variation

FLUX.1 Redux is the image-to-image variation model in the FLUX.1 Tools suite. Rather than ControlNet-style structural conditioning, Redux uses an image encoder (a SigLIP-style vision transformer) to extract a dense embedding of the input image, which is then injected into the FLUX.1 DiT alongside the text embeddings via cross-attention. This is conceptually similar to IP-Adapter and Stable Diffusion's 'unCLIP': the result is a new image that captures the style, palette and high-level composition of the input but is regenerated from noise, producing diverse variations. Optionally a text prompt can steer the variation toward a different style or composition. The base backbone is the standard FLUX.1 rectified-flow DiT with T5-XXL and CLIP-L text encoders and a 16x VAE, sampled in 28-50 flow-matching steps.

Parameters
~12B (FLUX.1 dev backbone) plus image-encoder adapter
Context
512 tokens
What it can do
  • Image-to-image variation conditioned on a reference image
  • Optional text prompt to steer style or content of variations
  • Preserves overall composition and palette while regenerating details
  • Up to 2 MP output
  • Compatible with FLUX Tools Fill/Canny/Depth pipelines
  • Good for unCLIP-style 'more like this' workflows
  • Best for: design exploration, A/B variants, mood-board expansion, style-transfer-lite.
Training & License

Fine-tuned from FLUX.1 base weights using paired references and image-text data, with a SigLIP-style image encoder trained or adapted to produce conditioning embeddings.

License: FLUX.1 Redux [dev] under the FLUX.1 [dev] non-commercial license; FLUX.1 Redux [pro] via API for commercial use.

Known limitations
  • Not pixel-faithful β€” output differs from reference
  • Cannot perform precise edits (use Fill for that)
  • Dev weights non-commercial only
  • Tends to homogenise unusual reference styles toward FLUX defaults

Frequently asked questions

Start using FLUX.1 Redux today

Get started with free credits. No credit card required. Access FLUX.1 Redux and 100+ other models through a single API.