### Start Qwen-Image Gradio Demo Server

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Starts the Gradio demo server for Qwen-Image, enabling a web interface with multi-GPU processing, queue management, and prompt optimization. Requires setting the DASHSCOPE_API_KEY.

```bash
cd src
DASHSCOPE_API_KEY=sk-xxxxxxxxxxxxxxxxx python examples/demo.py 
```

--------------------------------

### Run Prompt Enhancement Script from Command Line

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Executes the prompt enhancement example script from the command line. Requires setting the DASHSCOPE_API_KEY environment variable.

```bash
cd src
DASHSCOPE_API_KEY=sk-xxxxxxxxxxxxxxxxxxxx python examples/generate_w_prompt_enhance.py
```

--------------------------------

### Server Configuration and MultiGPUManager

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Example server configuration showing how environment variables are read and printed. The MultiGPUManager handles task distribution, queue management, progress tracking, and shutdown.

```python
import os
import torch
from multiprocessing import Process, Queue, Event

# Configuration from environment
NUM_GPUS_TO_USE = int(os.environ.get("NUM_GPUS_TO_USE", torch.cuda.device_count()))
TASK_QUEUE_SIZE = int(os.environ.get("TASK_QUEUE_SIZE", 100))
TASK_TIMEOUT = int(os.environ.get("TASK_TIMEOUT", 300))

print(f"Config: Using {NUM_GPUS_TO_USE} GPUs, queue size {TASK_QUEUE_SIZE}, timeout {TASK_TIMEOUT} seconds")

# The MultiGPUManager handles:
# - Task distribution across GPUs
# - Queue management for high concurrency
# - Progress tracking
# - Graceful shutdown with cleanup

# API available at:
# - GET /api/status - Returns queue and worker status
# - POST /api/generate - Submit generation task
# Output: JSON with task_queue_size, result_queue_size, pending_tasks, active_workers
```

--------------------------------

### Full Example: Command-Line Generation with Prompt Enhancement

Source: https://context7.com/qwenlm/qwen-image/llms.txt

A complete Python script demonstrating command-line image generation with automatic prompt enhancement using the Qwen-Image pipeline. Includes aspect ratio definitions and prompt rewriting.

```python
from diffusers import DiffusionPipeline
from tools.prompt_utils import rewrite
import torch

# Initialize the pipeline
pipe = DiffusionPipeline.from_pretrained("Qwen/Qwen-Image", torch_dtype=torch.bfloat16)
pipe = pipe.to("cuda")

# Available aspect ratios with optimal dimensions
aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1140),
    "3:4": (1140, 1472)
}

prompt = "一只可爱的小猫坐在花园里"  # Chinese prompt
prompt = rewrite(prompt)  # Enhance prompt automatically

width, height = aspect_ratios["16:9"]

image = pipe(
    prompt=prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device="cuda").manual_seed(42)
).images[0]

image.save("example.png")
```

--------------------------------

### Load and Use Qwen-Image-Edit-2509 Pipeline

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Loads the QwenImageEditPlusPipeline, processes two input images with a given prompt, and saves the output. Ensure you have the necessary libraries installed and a CUDA-enabled GPU for optimal performance.

```python
import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
from io import BytesIO
import requests

pipeline = QwenImageEditPlusPipeline.from_pretrained("Qwen/Qwen-Image-Edit-2509", torch_dtype=torch.bfloat16)
print("pipeline loaded")

pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)
image1 = Image.open(BytesIO(requests.get("https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/edit2509/edit2509_1.jpg").content))
image2 = Image.open(BytesIO(requests.get("https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/edit2509/edit2509_2.jpg").content))
prompt = "The magician bear is on the left, the alchemist bear is on the right, facing each other in the central park square."
inputs = {
    "image": [image1, image2],
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 40,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}
with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit_plus.png")
    print("image saved at", os.path.abspath("output_image_edit_plus.png"))
```

--------------------------------

### Install Diffusers Library

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Install the latest version of the diffusers library. Ensure your transformers version is at least 4.51.3.

```bash
pip install git+https://github.com/huggingface/diffusers
```

--------------------------------

### Multi-GPU API Server Deployment Configuration

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Configure environment variables for deploying Qwen-Image as a multi-GPU API server. Sets the number of GPUs, task queue size, and task timeout.

```bash
# Set environment variables for configuration
export NUM_GPUS_TO_USE=4          # Number of GPUs to use
export TASK_QUEUE_SIZE=100        # Task queue size
export TASK_TIMEOUT=300           # Task timeout in seconds
export DASHSCOPE_API_KEY=sk-xxxxxxxxxxxxxxxxx  # For prompt enhancement

# Start the Gradio demo server
cd src
python examples/demo.py
# Output: Server starts at http://0.0.0.0:7860 with multi-GPU support
```

--------------------------------

### Qwen-Image Pipeline for Text-to-Image Generation

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Illustrates text-to-image generation using the Qwen-Image DiffusionPipeline. It supports specifying different aspect ratios and includes positive/negative prompts. Ensure PyTorch is available and the device is set correctly.

```python
from diffusers import DiffusionPipeline
import torch

model_name = "Qwen/Qwen-Image"

# Load the pipeline
if torch.cuda.is_available():
    torch_dtype = torch.bfloat16
    device = "cuda"
else:
    torch_dtype = torch.float32
    device = "cpu"

pipe = DiffusionPipeline.from_pretrained(model_name, torch_dtype=torch_dtype).to(device)

positive_magic = {
    "en": ", Ultra HD, 4K, cinematic composition.", # for english prompt
    "zh": ", 超清，4K，电影级构图."
}

# Generate image
prompt = '''A coffee shop entrance features a chalkboard sign reading "Qwen Coffee 😊 $2 per cup," with a neon light beside it displaying "通义千问". Next to it hangs a poster showing a beautiful Chinese woman, and beneath the poster is written "π≈3.1415926-53589793-23846264-33832795-02384197".'''

negative_prompt = " " # Recommended if you don't use a negative prompt.


# Generate with different aspect ratios
aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1104),
    "3:4": (1104, 1472),
    "3:2": (1584, 1056),
    "2:3": (1056, 1584),
}

width, height = aspect_ratios["16:9"]

image = pipe(
    prompt=prompt + positive_magic["en"],
    negative_prompt=negative_prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device="cuda").manual_seed(42)
).images[0]

image.save("example.png")
```

--------------------------------

### Configure Multi-GPU API Server Environment Variables

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Sets environment variables to configure the Multi-GPU API Server for Qwen-Image, controlling the number of GPUs, task queue size, and task timeout.

```bash
export NUM_GPUS_TO_USE=4          # Number of GPUs to use
export TASK_QUEUE_SIZE=100        # Task queue size
export TASK_TIMEOUT=300           # Task timeout in seconds
```

--------------------------------

### Text to Image Generation with Qwen-Image-2512

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Load the QwenImagePipeline and generate an image from a detailed text prompt. Supports various aspect ratios and custom inference steps. Ensure CUDA is available for optimal performance.

```python
from diffusers import QwenImagePipeline
import torch
# Load the pipeline
if torch.cuda.is_available():
    torch_dtype = torch.bfloat16
    device = "cuda"
else:
    torch_dtype = torch.float32
    device = "cpu"

pipe = QwenImagePipeline.from_pretrained("Qwen/Qwen-Image-2512", torch_dtype=torch_dtype).to(device)

# Generate image
prompt = '''A 20-year-old East Asian girl with delicate, charming features and large, bright brown eyes—expressive and lively, with a cheerful or subtly smiling expression. Her naturally wavy long hair is either loose or tied in twin ponytails. She has fair skin and light makeup accentuating her youthful freshness. She wears a modern, cute dress or relaxed outfit in bright, soft colors—lightweight fabric, minimalist cut. She stands indoors at an anime convention, surrounded by banners, posters, or stalls. Lighting is typical indoor illumination—no staged lighting—and the image resembles a casual iPhone snapshot: unpretentious composition, yet brimming with vivid, fresh, youthful charm.'''

negative_prompt = "低分辨率，低画质，肢体畸形，手指畸形，画面过饱和，蜡像感，人脸无细节，过度光滑，画面具有AI感。构图混乱。文字模糊，扭曲。"


# Generate with different aspect ratios
aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1104),
    "3:4": (1104, 1472),
    "3:2": (1584, 1056),
    "2:3": (1056, 1584),
}

width, height = aspect_ratios["16:9"]

image = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device="cuda").manual_seed(42)
).images[0]

image.save("example.png")

```

--------------------------------

### Command-Line Generation with Prompt Enhancement

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Run image generation from the command line with automatic prompt enhancement. Requires DASHSCOPE_API_KEY to be set for prompt rewriting.

```bash
# Set API key for prompt enhancement
export DASHSCOPE_API_KEY=sk-xxxxxxxxxxxxxxxxxxxx

# Run the generation script
cd src
python examples/generate_w_prompt_enhance.py

# The script:
# 1. Loads Qwen-Image pipeline
# 2. Takes a simple prompt
# 3. Automatically enhances it via Qwen-Plus API
# 4. Generates and saves the image as example.png
```

--------------------------------

### Qwen-Image-Edit-2511 Pipeline for Image Editing

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Demonstrates how to use the QwenImageEditPlusPipeline for image editing with support for multiple image inputs and improved consistency. Ensure the model is loaded and moved to the appropriate device (e.g., 'cuda').

```python
import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
from io import BytesIO
import requests

pipeline = QwenImageEditPlusPipeline.from_pretrained("Qwen/Qwen-Image-Edit-2511", torch_dtype=torch.bfloat16)
print("pipeline loaded")

pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)
image1 = Image.open(BytesIO(requests.get("https://qianwen-res.oss-accelerate-overseas.aliyuncs.com/Qwen-Image/edit2511/edit2511input.png").content))
prompt = "这个女生看着面前的电视屏幕，屏幕上面写着“阿里巴巴”"
inputs = {
    "image": [image1],
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 40,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}
with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit_2511.png")
    print("image saved at", os.path.abspath("output_image_edit_2511.png"))
```

--------------------------------

### Qwen-Image-Edit Pipeline for Single Image Editing

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Shows how to use the QwenImageEditPipeline for editing a single image. It's recommended to use Qwen-Image-Edit-2509 for better consistency. Prompt rewriting is strongly recommended for stable editing results.

```python
import os
from PIL import Image
import torch

from diffusers import QwenImageEditPipeline

pipeline = QwenImageEditPipeline.from_pretrained("Qwen/Qwen-Image-Edit")
print("pipeline loaded")
pipeline.to(torch.bfloat16)
pipeline.to("cuda")
pipeline.set_progress_bar_config(disable=None)

image = Image.open("./input.png").convert("RGB")
prompt = "Change the rabbit's color to purple, with a flash light background."


inputs = {
    "image": image,
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 50,
}

with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit.png")
    print("image saved at", os.path.abspath("output_image_edit.png"))
```

--------------------------------

### Integrate Prompt Enhancement for Text-to-Image

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Integrates the official Prompt Enhancement Tool for optimized prompts and multi-language support in text-to-image generation. Requires the 'tools.prompt_utils' module.

```python
from tools.prompt_utils import rewrite
prompt = rewrite(prompt)
```

--------------------------------

### Load and Use QwenImageEditPlusPipeline for Single Image Editing

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Loads the QwenImageEditPlusPipeline for single image editing. Ensure the model name and torch_dtype are correctly specified. The pipeline is moved to CUDA for accelerated processing.

```python
pipeline = QwenImageEditPlusPipeline.from_pretrained(
    "Qwen/Qwen-Image-Edit-2511",
    torch_dtype=torch.bfloat16
)
print("pipeline loaded")

pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)

# Load input image
image1 = Image.open(BytesIO(requests.get(
    "https://qianwen-res.oss-accelerate-overseas.aliyuncs.com/Qwen-Image/edit2511/edit2511input.png"
).content))

# Define edit instruction (supports Chinese and English)
prompt = "这个女生看着面前的电视屏幕，屏幕上面写着"阿里巴巴""

# Configure generation parameters
inputs = {
    "image": [image1],
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 40,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}

# Generate edited image
with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit_2511.png")
    print("image saved at", os.path.abspath("output_image_edit_2511.png"))
```

--------------------------------

### Prompt Enhancement for Text-to-Image Generation

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Utilizes the prompt rewriting utility powered by Qwen-Plus to optimize prompts for text-to-image generation. The utility automatically detects language and applies enhancement strategies. Requires initializing the DiffusionPipeline and moving it to CUDA.

```python
from tools.prompt_utils import rewrite
from diffusers import DiffusionPipeline
import torch

# Initialize the pipeline
pipe = DiffusionPipeline.from_pretrained("Qwen/Qwen-Image", torch_dtype=torch.bfloat16)
pipe = pipe.to("cuda")

# Available aspect ratios
aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1140),
    "3:4": (1140, 1472)
}

# Original simple prompt (Chinese)
prompt = "一只可爱的小猫坐在花园里"
```

--------------------------------

### Text-to-Image Generation with Original Qwen-Image Model

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Generates images using the original Qwen-Image model via the DiffusionPipeline. This method supports basic text-to-image generation and various aspect ratios. Magic suffixes can enhance prompt quality.

```python
from diffusers import DiffusionPipeline
import torch

model_name = "Qwen/Qwen-Image"

# Load the pipeline
if torch.cuda.is_available():
    torch_dtype = torch.bfloat16
    device = "cuda"
else:
    torch_dtype = torch.float32
    device = "cpu"

pipe = DiffusionPipeline.from_pretrained(model_name, torch_dtype=torch_dtype).to(device)

# Magic suffixes to enhance prompt quality
positive_magic = {
    "en": ", Ultra HD, 4K, cinematic composition.",
    "zh": ", 超清，4K，电影级构图."
}

# Generate image with text rendering
prompt = '''A coffee shop entrance features a chalkboard sign reading "Qwen Coffee 😊 $2 per cup," with a neon light beside it displaying "通义千问". Next to it hangs a poster showing a beautiful Chinese woman, and beneath the poster is written "π≈3.1415926-53589793-23846264-33832795-02384197".'''

negative_prompt = " "  # Recommended if you don't use a negative prompt

aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1104),
    "3:4": (1104, 1472),
}

width, height = aspect_ratios["16:9"]

image = pipe(
    prompt=prompt + positive_magic["en"],
    negative_prompt=negative_prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device="cuda").manual_seed(42)
).images[0]

image.save("coffee_shop.png")
# Output: Image with accurately rendered English, Chinese text, and mathematical symbols
```

--------------------------------

### Text-to-Image Generation with Qwen-Image-2512

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Generates high-quality images from text prompts using the Qwen-Image-2512 model. Load the pipeline with bfloat16 for GPU efficiency and specify desired aspect ratios. Ensure CUDA is available for optimal performance.

```python
from diffusers import QwenImagePipeline
import torch

# Load the pipeline with bfloat16 for GPU efficiency
if torch.cuda.is_available():
    torch_dtype = torch.bfloat16
    device = "cuda"
else:
    torch_dtype = torch.float32
    device = "cpu"

pipe = QwenImagePipeline.from_pretrained("Qwen/Qwen-Image-2512", torch_dtype=torch_dtype).to(device)

# Define prompt and negative prompt
prompt = '''A 20-year-old East Asian girl with delicate, charming features and large, bright brown eyes—expressive and lively, with a cheerful or subtly smiling expression. Her naturally wavy long hair is either loose or tied in twin ponytails. She has fair skin and light makeup accentuating her youthful freshness.'''

negative_prompt = "低分辨率，低画质，肢体畸形，手指畸形，画面过饱和，蜡像感，人脸无细节，过度光滑，画面具有AI感。构图混乱。文字模糊，扭曲。"

# Available aspect ratios with optimal dimensions
aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1104),
    "3:4": (1104, 1472),
    "3:2": (1584, 1056),
    "2:3": (1056, 1584),
}

width, height = aspect_ratios["16:9"]

# Generate the image
image = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device="cuda").manual_seed(42)
).images[0]

image.save("generated_image.png")
# Output: A high-quality 1664x928 image saved to generated_image.png
```

--------------------------------

### Integrate Prompt Enhancement for Image Edit

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Integrates the official Prompt Enhancement Tool for enhanced stability in image editing tasks. Requires the 'tools.prompt_utils' module and a PIL image object.

```python
from tools.prompt_utils import polish_edit_prompt
prompt = polish_edit_prompt(prompt, pil_image)
```

--------------------------------

### Rewrite Prompt for Enhanced Quality

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Rewrite a given prompt to improve image generation quality. Requires the DASHSCOPE_API_KEY environment variable to be set.

```python
# export DASHSCOPE_API_KEY=sk-xxxxxxxxxxxxxxxxxxxx
prompt = rewrite(prompt)
print(f"Enhanced prompt: {prompt}")

width, height = aspect_ratios["16:9"]

image = pipe(
    prompt=prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device="cuda").manual_seed(42)
).images[0]

image.save("example.png")
# Output: High-quality image generated from enhanced prompt with additional visual details
```

--------------------------------

### Generate Image with SGLang

Source: https://github.com/qwenlm/qwen-image/blob/main/README.md

Use this command to generate or edit images with Qwen-Image using SGLang. Specify the model path, prompt, and image paths for editing tasks.

```bash
sglang generate --model-path Qwen/Qwen-Image-Edit-2511 --prompt "make the girl in Figure 1 dance with the capybara in Figure 2."  --image-path "https://github.com/lm-sys/lm-sys.github.io/releases/download/test/TI2I_Qwen_Image_Edit_Input.jpg" "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/edit2509/edit2509_2.jpg"
```

--------------------------------

### Prompt Enhancement for Image Editing

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Enhance edit prompts using the vision-language model for more specific editing instructions based on image context. Requires DASH_API_KEY environment variable.

```python
from tools.prompt_utils import polish_edit_prompt
from PIL import Image
from diffusers import QwenImageEditPipeline
import torch

# Load the editing pipeline
pipeline = QwenImageEditPipeline.from_pretrained("Qwen/Qwen-Image-Edit")
pipeline.to(torch.bfloat16)
pipeline.to("cuda")

# Load input image
pil_image = Image.open("./portrait.png").convert("RGB")

# Original simple edit instruction
original_prompt = "Change the person's hat"

# Enhance the edit prompt (requires DASH_API_KEY environment variable)
# export DASH_API_KEY=sk-xxxxxxxxxxxxxxxxxxxx
enhanced_prompt = polish_edit_prompt(original_prompt, pil_image)
print(f"Enhanced prompt: {enhanced_prompt}")
# Example output: "Replace the man's hat with a dark brown beret; keep smile, short hair, and gray jacket unchanged"

# Generate edited image with enhanced prompt
inputs = {
    "image": pil_image,
    "prompt": enhanced_prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 50,
}

with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("edited_portrait.png")
# Output: More accurate edit with preserved identity features
```

--------------------------------

### Multi-Image Editing with Qwen-Image-Edit-2509

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Combines multiple input images into a single output using the Qwen-Image-Edit-2509 model. Supports various combinations like person+person, person+scene, and person+object with up to 3 images.

```python
import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
from io import BytesIO
import requests

# Load the multi-image editing pipeline
pipeline = QwenImageEditPlusPipeline.from_pretrained(
    "Qwen/Qwen-Image-Edit-2509",
    torch_dtype=torch.bfloat16
)
print("pipeline loaded")

pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)

# Load multiple input images
image1 = Image.open(BytesIO(requests.get(
    "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/edit2509/edit2509_1.jpg"
).content))
image2 = Image.open(BytesIO(requests.get(
    "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/edit2509/edit2509_2.jpg"
).content))

# Describe how to combine the images
prompt = "The magician bear is on the left, the alchemist bear is on the right, facing each other in the central park square."

inputs = {
    "image": [image1, image2],  # List of input images
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 40,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}

with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit_plus.png")
    print("image saved at", os.path.abspath("output_image_edit_plus.png"))
```

--------------------------------

### Single Image Editing with Qwen-Image-Edit

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Performs single-image editing operations like style transfer, object modification, text editing, and background changes using the original Qwen-Image-Edit model. The image is loaded and converted to RGB before processing.

```python
import os
from PIL import Image
import torch
from diffusers import QwenImageEditPipeline

# Load the single-image editing pipeline
pipeline = QwenImageEditPipeline.from_pretrained("Qwen/Qwen-Image-Edit")
print("pipeline loaded")
pipeline.to(torch.bfloat16)
pipeline.to("cuda")
pipeline.set_progress_bar_config(disable=None)

# Load input image
image = Image.open("./input.png").convert("RGB")

# Define edit instruction
prompt = "Change the rabbit's color to purple, with a flash light background."

inputs = {
    "image": image,
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 50,
}

with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit.png")
    print("image saved at", os.path.abspath("output_image_edit.png"))
```

--------------------------------

### Image Editing with Qwen-Image-Edit-2511

Source: https://context7.com/qwenlm/qwen-image/llms.txt

Edits existing images using natural language instructions with the Qwen-Image-Edit-2511 model. This model supports multi-image input, improved character consistency, and built-in LoRA capabilities.

```python
import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
from io import BytesIO
import requests
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.