Home › Image Models › FLUX.1-dev

FLUX.1-dev - AI Image Models Tool

Overview

FLUX.1-dev is a Diffusers- and ComfyUI-compatible text-to-image model published on Hugging Face by black-forest-labs. The model is presented as a high-quality image generation checkpoint intended for creative, non-commercial projects; its Hugging Face model page lists the license as “other,” indicating a non-commercial restriction. FLUX.1-dev has attracted substantial community attention on Hugging Face, with hundreds of thousands of downloads and more than ten thousand likes, which suggests broad experimentation and adoption among hobbyists and researchers. The model is distributed as a pipeline-ready checkpoint (text-to-image) and explicitly supports both ComfyUI node-based workflows and the Hugging Face Diffusers ecosystem, making it easy to integrate into existing generative-image pipelines. Because the model card does not publish parameter counts or a formal benchmark suite, users should rely on empirical testing for throughput, memory use, and quality tradeoffs on their target hardware. For the latest release notes, compatibility details, and usage examples, refer to the model repository on Hugging Face (black-forest-labs/FLUX.1-dev).

Model Statistics

Downloads: 674,581
Likes: 12131
Pipeline: text-to-image

License: other

Model Details

Architecture and compatibility: FLUX.1-dev is provided as a text-to-image checkpoint compatible with the Hugging Face Diffusers pipeline and with ComfyUI node graphs. The model card and repository list the pipeline type as text-to-image, but do not disclose a base model, parameter count, or detailed architecture diagram. According to the Hugging Face page, the model uses a non-commercial (“other”) license (see model repository). Capabilities and integration: Because the model is Diffusers-compatible, standard Diffusers features—such as choosing schedulers, switching samplers, using guidance scales, and applying attention-slicing/CPU offload strategies—can typically be used when running the checkpoint in that ecosystem. ComfyUI compatibility enables visual node-based prompt workflows, iterative refinement, and integration with other image-processing nodes. Users should validate sampler compatibility and recommended inference settings on their hardware; the model card does not publish official throughput or latency numbers. Community notes: The repository’s Hugging Face page shows substantial downloads (674,581) and likes (12,131), indicating active community usage and experimentation. For up-to-date changelogs or specific engineering details, consult the model repo and its README on Hugging Face.

Key Features

High-quality text-to-image generation optimized for creative outputs
Direct compatibility with Hugging Face Diffusers pipelines
Supports ComfyUI node-based visual workflows
Distributed under a non-commercial ("other") license
Large community adoption on Hugging Face (downloads and likes)

Example Usage

Example (python):

from diffusers import DiffusionPipeline
import torch

model_id = "black-forest-labs/FLUX.1-dev"

def generate(prompt, out_path="flux_output.png", device=None):
    device = device or ("cuda" if torch.cuda.is_available() else "cpu")
    # Load pipeline; adjust torch_dtype and safety checker per your environment
    pipe = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16 if device=="cuda" else torch.float32)
    pipe = pipe.to(device)

    # Example inference settings; tune guidance_scale, num_inference_steps, and seed
    image = pipe(prompt, guidance_scale=7.5, num_inference_steps=30).images[0]
    image.save(out_path)
    print(f"Saved image to {out_path}")

if __name__ == "__main__":
    example_prompt = "A cinematic portrait of a futuristic forest city at golden hour, ultra-detailed"
    generate(example_prompt)