FLUX.1 and Fooocus Guide: Compatible UIs and ComfyUI Setup

Fooocus is an image-generation interface built around SDXL, while FLUX.1 is a separate image-model family from Black Forest Labs. The official Fooocus project does not provide native FLUX support and directs users interested in newer architectures to Forge, ComfyUI/SwarmUI, or community forks. To generate with FLUX, use one of those explicitly compatible implementations and obtain the appropriate model weights. Fooocus README, FLUX announcement

For a Fooocus user, start by choosing an interface with documented FLUX support, then complete its supplied example workflow before rebuilding your usual presets.

What are the key facts about Fooocus and FLUX.1?

Which interfaces support a Fooocus user's move to FLUX?

Fooocus organizes generation around prompts and images, with built-in paths for variations, upscaling, inpainting, and image prompts. Those features explain why its interface can be appealing for an SDXL workflow. Its README also says the project is in limited long-term support focused on bug fixes. Fooocus README

For FLUX, ComfyUI provides official workflows covering dev and schnell, including separate component downloads and packaged FP8 checkpoints. The graph makes the diffusion model, encoders, and VAE visible as selected resources. This is useful when learning what a local FLUX pipeline actually loads. ComfyUI tutorial

Forge provides another documented option. Its README includes FLUX tutorials, supported quantized formats, and memory-management controls. Choose it when a WebUI-style interface is more comfortable than a graph of connected nodes. That is a workflow preference, not a quality ranking between interfaces. Forge README

There are also independently maintained Fooocus derivatives. RuinedFooocus, for example, explicitly lists FLUX support in its README. That feature belongs to that project, so evaluate its own instructions and releases rather than assuming it changes the capabilities of official Fooocus. RuinedFooocus README

What are the limits when moving from Fooocus to FLUX?

The Fooocus maintainers state that there are no current plans to incorporate newer model architectures. Moving a FLUX file into Fooocus’s checkpoint folder therefore does not constitute a documented installation method. Use the supported interface rather than treating the file extension as proof of compatibility. Fooocus README

Do not transfer hardware expectations from Fooocus directly to FLUX. The applications may have similar controls, but the model and loaded components differ. Diffusers warns that FLUX can be costly to run on consumer hardware and discusses memory optimizations separately. Diffusers FLUX documentation

Model licenses also remain distinct. FLUX.1-schnell declares Apache-2.0, while FLUX.1-dev uses its own Non-Commercial License. An interface being free to download does not replace the license of the weights loaded into it. Schnell card, Dev card

For an initial comparison, carry over the scene description and intended composition. Treat prompt weighting syntax, negative prompts, adapters, and saved presets as implementation-specific until the destination workflow documents them. This is a migration recommendation, not a claim that every familiar control has an equivalent.

How can a Fooocus user generate FLUX images in ComfyUI?

The following route uses ComfyUI’s official FLUX.1-schnell example. Start with a working, updated ComfyUI installation, then download the workflow image from the official tutorial and drag it into the interface. The embedded workflow provides the graph and model information. ComfyUI tutorial

Follow the tutorial’s manual installation section if model downloads are not handled automatically. The full schnell workflow uses flux1-schnell.safetensors, clip_l.safetensors, a T5 encoder, and ae.safetensors. Keep each file in its documented model directory. ComfyUI tutorial

With Hugging Face’s hf CLI installed and any repository access requirements completed, this command downloads the diffusion model from the ComfyUI installation directory:

hf download black-forest-labs/FLUX.1-schnell \
  flux1-schnell.safetensors \
  --local-dir ./models/diffusion_models

This command downloads only the named diffusion-model file. Obtain the encoders and VAE from the tutorial’s links as well. Hugging Face documents the CLI’s repository, filename, and destination-directory arguments. CLI guide, ComfyUI tutorial

Select the matching files in the workflow’s model loaders. Use the supplied schnell workflow settings, enter a concrete scene such as “A green bicycle leaning against a bakery wall in morning light,” and run the graph. Save the first successful result before adjusting the workflow.

Schnell’s model card describes generation in one to four sampling steps, and its reference Python example uses four. Those are sampling steps, not a promise of completion in four seconds. Let the workflow supply its matching settings instead of copying a Fooocus performance preset. Schnell model card

If the graph reports missing models, compare the selected filenames against the tutorial before downloading alternatives. If generation fails for memory reasons, review the documented FP8 workflow or supported optimization options. Avoid changing the base model, encoders, and precision simultaneously while diagnosing the first run. ComfyUI tutorial, Diffusers documentation

Once the baseline works, compare a few scenes you already understand from Fooocus. Keep a note of which differences you prefer and which controls you miss. That gives you a practical basis for choosing a working environment.

How do Fooocus, Forge, and ComfyUI compare?

For the SDXL workflow, see the Fooocus complete guide. The sibling FLUX guide for AUTOMATIC1111 users covers the Forge migration path.

What else should Fooocus users know about FLUX?

Does official Fooocus support FLUX?

Official Fooocus uses the SDXL architecture and directs FLUX users to other applications. Its README recommends Forge, ComfyUI/SwarmUI, or community forks for newer models. Fooocus README

Is RuinedFooocus the same as official Fooocus?

RuinedFooocus is an independently maintained derivative whose feature list includes FLUX support. That feature belongs to RuinedFooocus and does not establish support in official Fooocus. RuinedFooocus README, Fooocus README

Which FLUX model does this ComfyUI workflow use?

The workflow above uses FLUX.1-schnell with ComfyUI's supplied example graph. Schnell documents generation in one to four sampling steps and uses the Apache-2.0 license. ComfyUI tutorial, Schnell card

Can I reuse Fooocus LoRAs in a FLUX workflow?

Check each LoRA's declared base model before loading it into a FLUX workflow. Official Fooocus is SDXL-based, while ComfyUI's FLUX examples load FLUX-specific components, so the model bases must be checked separately. Fooocus README, ComfyUI tutorial

Sources

• Official Fooocus README
• BFL’s FLUX launch announcement
• FLUX.1-dev model card
• FLUX.1-schnell model card
• ComfyUI FLUX workflow
• Forge README
• RuinedFooocus README
• Diffusers FLUX documentation
• Hugging Face CLI guide

Related guides on PromptZone

• Fooocus 2026: The Complete Guide
• The Ultimate Guide to Fooocus Image Prompts
• Mastering Fooocus Inpainting