InstantX / FLUX.1-dev-Controlnet-Union

FLUX.1-dev Controlnet

Diffusers version

until the next Diffusers pypi release, please install Diffusers from source and use this PR to be able to use. TODO: change when new version.

Checkpoint

The training of union controlnet requires a significant amount of computational power. The current release is only an alpha version checkpoint that has not been fully trained. The beta version is in the training process. We have conducted ablation studies that have demonstrated the validity of the code. The open-source release of the alpha version is solely to facilitate the rapid growth of the open-source community and the Flux ecosystem; it is common to encounter bad cases (please accept my apologies). It is worth noting that we have found that even a fully trained Union model may not perform as well as specialized models, such as pose control. However, as training progresses, the performance of the Union model will continue to approach that of specialized models.

Control Mode

Demo

import torch
from diffusers.utils import load_image
from diffusers.pipelines.flux.pipeline_flux_controlnet import FluxControlNetPipeline
from diffusers.models.controlnet_flux import FluxControlNetModel

# load
base_model = 'black-forest-labs/FLUX.1-dev'
controlnet_model = 'InstantX/FLUX.1-dev-Controlnet-Union-alpha'
controlnet = FluxControlNetModel.from_pretrained(controlnet_model, torch_dtype=torch.bfloat16)
pipe = FluxControlNetPipeline.from_pretrained(base_model, controlnet=controlnet, torch_dtype=torch.bfloat16)
pipe.to("cuda")

# image cfg
width, height = 1024, 1024
controlnet_conditioning_scale = 0.5
seed = 6666

# canny
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/canny.jpg")
prompt = "A girl in city, 25 years old, cool, futuristic."
control_mode = 0

# tile
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/tile.jpg")
prompt = "A girl, 25 years old."
control_mode = 1

# depth
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/depth.jpg")
prompt = "A girl in city, 25 years old, cool, futuristic."
control_mode = 2

# blur
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/blur.jpg")
prompt = "A girl, 25 years old."
control_mode = 3

# pose
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/pose.jpg")
prompt = "A girl in city, 25 years old, cool, futuristic."
control_mode = 4

# gray
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/gray.jpg")
prompt = "A girl, 25 years old."
control_mode = 5

# low quality
control_image = load_image("https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Union-alpha/resolve/main/images/lq.jpg")
prompt = "A girl in city"
control_mode = 6

# go go go
image = pipe(
    prompt, 
    control_image=control_image,
    control_mode=control_mode,
    width=width,
    height=height,
    controlnet_conditioning_scale=controlnet_conditioning_scale,
    num_inference_steps=28, 
    guidance_scale=3.5,
    generator=torch.manual_seed(seed),
).images[0]
image.save("image.jpg")

Acknowledgements

Thank you, zzzzzero , for pointing out the bug in the model.