Featured image of post ComfyUI 整合 Qwen Image 量化版

ComfyUI 整合 Qwen Image 量化版

阿里巴巴达摩院发布了文生图模型 Qwen-Image,该模型基于拥有 200 亿参数的 MMDiT 架构,具备生成写实、动漫等数十种图像风格的能力,并支持风格迁移等常见图像编辑操作

ComfyUI 整合 Qwen Image 量化版

模型下载

可使用 hf-mirror.com 进行国内加速下载,例如 https://hf-mirror.com/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors

qwen_image_fp8_e4m3fn.safetensors

下载放置 ComfyUI/models/diffusion_models 目录中

1
https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors

qwen_2.5_vl_7b_fp8_scaled.safetensors

下载放置 ComfyUI/models/text_encoders 目录中

1
https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors

qwen_image_vae.safetensors

下载放置 ComfyUI/models/vae 目录中

1
https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors

模型目录大概如下所示:

📂 ComfyUI/

├── 📂 models/

│ ├── 📂 diffusion_models/

│ │ └── qwen_image_fp8_e4m3fn.safetensors

│ ├── 📂 vae/

│ │ └── qwen_image_vae.safetensors

│ └── 📂 text_encoders/

│ └── qwen_2.5_vl_7b_fp8_scaled.safetensors

工作流下载

下载工作流并导入到 ComfyUI

1
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/refs/heads/main/templates/image_qwen_image.json

qwen_image_01.png

左侧的模型配置,对应的选择上面下载好的模型位置即可

测试

直接用官方的工作流试试效果,里面已经带有一些中文文字

如果运行时候提示 “Prompt outputs failed validation: CLIPLoader: - Value not in list: type: ‘qwen_image’ not in [‘stable_diffusion’, ‘stable_cascade’, ‘sd3’, ‘stable_audio’, ‘mochi’, ’ltxv’, ‘pixart’, ‘cosmos’, ’lumina2’, ‘wan’, ‘hidream’, ‘chroma’, ‘ace’, ‘omnigen2’]",先更新一下 ComfyUI 的版本

1
2
"A vibrant, warm neon-lit street scene in Hong Kong at the afternoon, with a mix of colorful Chinese and English signs glowing brightly. The atmosphere is lively, cinematic, and rain-washed with reflections on the pavement. The colors are vivid, full of pink, blue, red, and green hues. Crowded buildings with overlapping neon signs. 1980s Hong Kong style. Signs include:
"龍鳳冰室" "金華燒臘" "HAPPY HAIR" "鴻運茶餐廳" "EASY BAR" "永發魚蛋粉" "添記粥麵" "SUNSHINE MOTEL" "美都餐室" "富記糖水" "太平館" "雅芳髮型屋" "STAR KTV" "銀河娛樂城" "百樂門舞廳" "BUBBLE CAFE" "萬豪麻雀館" "CITY LIGHTS BAR" "瑞祥香燭莊" "文記文具" "GOLDEN JADE HOTEL" "LOVELY BEAUTY" "合興百貨" "興旺電器" And the background is warm yellow street and with all stores' lights on.

使用腾讯的 Cloud Studio 配置(16G 显存 + 32G 内存),跑的 1024 * 526,都要等很久,而且显存基本上占满了

qwen_image_02.png

效果如下图,量化版有些中文字也没有很好的表现出来

qwen_image_03.png

再来试试一个有中英文的场景,第二次生成就快了一些,效果如下图

qwen_image_04.png

Licensed under CC BY-NC-SA 4.0
本博客所有内容无特殊标注均为大卷学长原创内容,复制请保留原文出处。
Built with Hugo
Theme Stack designed by Jimmy