floyo logo
Powered by
ThinkDiffusion
floyo logo
Powered by
ThinkDiffusion
Now Available on Floyo

Flux 2 - Run in the Browser

Black Forest Labs' next-generation image model. Multi-reference consistency, readable text rendering, and sharper detail at higher resolutions. Run directly in your browser, no setup required.

Create with Flux 2
4MP
Max Edit Resolution
10
Reference Images
32B
Parameters (dev)
4
Model Variants

1. What's new in Flux 2

Improvements over Flux 1 focused on production reliability and multi-image workflows.

Multi-Reference Generation

Use up to 10 reference images at once for consistent characters, products, and visual style. Generate a character across multiple scenes and they'll look like the same person throughout.

Improved Detail & Photorealism

Sharper textures, more precise detailing, and stable lighting. Product shots look photographed, not generated. Fine details hold up when you zoom in.

Advanced Text Rendering

Typography, infographics, memes, and UI mockups generate with readable text. Small text renders legibly—no more garbled letters breaking the illusion.

Stronger Prompt Obedience

Complex prompts with multiple elements actually get interpreted accurately. Specify lighting, composition, and subject matter together without the model ignoring half your instructions.

2. Use cases

Where Flux 2's improvements make a real difference in production work.

Product Photography

Detail retention at higher resolutions means hero shots that don't fall apart at full size. Keep product appearance consistent across scenes.

Character Work

Storyboarding, concept art, social content. Multi-reference solves the "character looks different every time" problem.

Graphics with Text

Posters, social graphics, UI mockups. If your image needs readable words, Flux 2 handles it better than most alternatives.

Brand Consistency

Reference-based generation keeps brand elements, logos, and visual identity consistent across campaign assets.

Batch Production

Stronger prompt adherence means less regeneration hunting for outputs that match what you asked for.

High-Res Editing

Edit existing images at up to 4 megapixels while preserving clarity and coherence throughout.

3. Version comparison

Key improvements from Flux 1 to Flux 2.

Capability Flux 1 Flux 2
Multi-reference images Limited Up to 10
Text rendering Often garbled Clean & readable
Edit resolution Lower Up to 4MP
Complex prompts Partial Accurate
Detail consistency Good Excellent
Browser access ✓ Floyo

4. Model variants

Flux 2 comes in several versions for different needs.

Flux 2 [pro]

Full production-grade API. Best quality, fastest generation, zero compromises. This is what you're accessing through Floyo.

Flux 2 [flex]

Manual control over step count and guidance scale. Fine-tune quality vs. speed tradeoffs for your specific workflow.

Flux 2 [dev]

32B open-weight model. Run locally on RTX cards with fp8 optimization through ComfyUI. Weights available on Hugging Face.

Flux 2 [klein]

Coming soon. Apache 2.0 licensed, distilled for easier local deployment on consumer hardware.

5. Technical specs

Detailed specifications and parameters.

Access Browser-based (Floyo)
Edit Resolution Up to 4 megapixels
Reference Images Up to 10
Dev Model Size 32B parameters
Local Inference RTX GPUs (fp8)
Open Weights Hugging Face

6. How it works on Floyo

Understanding the workflow and what makes browser access practical.

Flux 2 is Black Forest Labs' follow-up to Flux 1, which became the most popular open-weight image model worldwide. Where Flux 1 demonstrated how powerful image models can be as creative tools, Flux 2 is built for actual production pipelines—not just impressive demos.

The main advantage of running Flux 2 through Floyo is skipping the local setup entirely. No downloading multi-gigabyte model weights, no VRAM management, no ComfyUI installation troubleshooting. Open your browser, load a workflow, start generating.

The ComfyUI interface is the same one you'd use locally, so workflows and node setups from tutorials translate directly. For multi-reference generation, connect your reference images to the appropriate input nodes and the workflow handles conditioning them properly so outputs stay consistent.

Flux 2 also includes a newly engineered VAE (variational autoencoder) that improves compression efficiency and output quality. This underpins all the flow backbones and contributes to the sharper detail at higher resolutions.

Start generating

Production-ready image generation directly in your browser. No setup, no installation, no API configuration.

Create with Flux 2

Free to try · No installation · Runs in the browser

Table of Contents