Qwen-Image controlnet inpaint + controlnet canny

[xinli2008]

Hey, Is it possible to load controlnet-canny and controlnet inpaint to inference a image?
Thanks

[mi804]

Yes. Please parse multiple controlnet modelconfig and call pipeline with multiple ControlNetInput with different controlnet_id

import torch
from PIL import Image
from modelscope import dataset_snapshot_download
from diffsynth.pipelines.qwen_image import QwenImagePipeline, ModelConfig, ControlNetInput


pipe = QwenImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="text_encoder/model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
        ModelConfig(model_id="DiffSynth-Studio/Qwen-Image-Blockwise-ControlNet-Inpaint", origin_file_pattern="model.safetensors"),
        ModelConfig(model_id="DiffSynth-Studio/Qwen-Image-Blockwise-ControlNet-Canny", origin_file_pattern="model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="tokenizer/"),
)

dataset_snapshot_download(
    dataset_id="DiffSynth-Studio/example_image_dataset",
    local_dir="./data/example_image_dataset",
    allow_file_pattern="canny/*.jpg"
)
controlnet_canny_image = Image.open("data/example_image_dataset/canny/image_1.jpg").resize((1328, 1328))

prompt = "一只小狗，毛发光洁柔顺，眼神灵动，背景是樱花纷飞的春日庭院，唯美温馨。"
controlnet_image = Image.open("./data/example_image_dataset/canny/image_2.jpg").convert("RGB").resize((1328, 1328))
# generate a centered square mask
inpaint_mask = Image.new("L", controlnet_image.size, 0)
mask_size = 512
left = (controlnet_image.width - mask_size) // 2
top = (controlnet_image.height - mask_size) // 2
right = left + mask_size
bottom = top + mask_size
inpaint_mask.paste(255, (left, top, right, bottom))
inpaint_mask = inpaint_mask.resize((1328, 1328)).convert("RGB")
inpaint_mask.save("inpaint_mask.jpg")
image = pipe(
    prompt, seed=0,
    input_image=controlnet_image, inpaint_mask=inpaint_mask,
    blockwise_controlnet_inputs=[
        ControlNetInput(image=controlnet_image, inpaint_mask=inpaint_mask, controlnet_id=0),
        ControlNetInput(image=controlnet_canny_image, controlnet_id=1)],
    num_inference_steps=40,
)
image.save("image.jpg")

[xinli2008]

Thanks, i will try.