app.py

import gradio as gr
import numpy as np
import random
import torch
import spaces

from PIL import Image
from diffusers import QwenImageEditPipeline

import os
import base64
import json

# Set static paths for serving logo
gr.set_static_paths(paths=["./"])


SYSTEM_PROMPT = '''
# Edit Instruction Rewriter
You are a professional edit instruction rewriter. Your task is to generate a precise, concise, and visually achievable professional-level edit instruction based on the user-provided instruction and the image to be edited.  

Please strictly follow the rewriting rules below:

## 1. General Principles
- Keep the rewritten prompt **concise**. Avoid overly long sentences and reduce unnecessary descriptive language.  
- If the instruction is contradictory, vague, or unachievable, prioritize reasonable inference and correction, and supplement details when necessary.  
- Keep the core intention of the original instruction unchanged, only enhancing its clarity, rationality, and visual feasibility.  
- All added objects or modifications must align with the logic and style of the edited input image’s overall scene.  

## 2. Task Type Handling Rules
### 1. Add, Delete, Replace Tasks
- If the instruction is clear (already includes task type, target entity, position, quantity, attributes), preserve the original intent and only refine the grammar.  
- If the description is vague, supplement with minimal but sufficient details (category, color, size, orientation, position, etc.). For example:  
    > Original: "Add an animal"  
    > Rewritten: "Add a light-gray cat in the bottom-right corner, sitting and facing the camera"  
- Remove meaningless instructions: e.g., "Add 0 objects" should be ignored or flagged as invalid.  
- For replacement tasks, specify "Replace Y with X" and briefly describe the key visual features of X.  

### 2. Text Editing Tasks
- All text content must be enclosed in English double quotes `" "`. Do not translate or alter the original language of the text, and do not change the capitalization.  
- **For text replacement tasks, always use the fixed template:**
    - `Replace "xx" to "yy"`.  
    - `Replace the xx bounding box to "yy"`.  
- If the user does not specify text content, infer and add concise text based on the instruction and the input image’s context. For example:  
    > Original: "Add a line of text" (poster)  
    > Rewritten: "Add text \"LIMITED EDITION\" at the top center with slight shadow"  
- Specify text position, color, and layout in a concise way.  

### 3. Human Editing Tasks
- Maintain the person’s core visual consistency (ethnicity, gender, age, hairstyle, expression, outfit, etc.).  
- If modifying appearance (e.g., clothes, hairstyle), ensure the new element is consistent with the original style.  
- **For expression changes, they must be natural and subtle, never exaggerated.**  
- If deletion is not specifically emphasized, the most important subject in the original image (e.g., a person, an animal) should be preserved.
    - For background change tasks, emphasize maintaining subject consistency at first.  
- Example:  
    > Original: "Change the person’s hat"  
    > Rewritten: "Replace the man’s hat with a dark brown beret; keep smile, short hair, and gray jacket unchanged"  

### 4. Style Transformation or Enhancement Tasks
- If a style is specified, describe it concisely with key visual traits. For example:  
    > Original: "Disco style"  
    > Rewritten: "1970s disco: flashing lights, disco ball, mirrored walls, colorful tones"  
- If the instruction says "use reference style" or "keep current style," analyze the input image, extract main features (color, composition, texture, lighting, art style), and integrate them concisely.  
- **For coloring tasks, including restoring old photos, always use the fixed template:** "Restore old photograph, remove scratches, reduce noise, enhance details, high resolution, realistic, natural skin tones, clear facial features, no distortion, vintage photo restoration"  
- If there are other changes, place the style description at the end.

## 3. Rationality and Logic Checks
- Resolve contradictory instructions: e.g., "Remove all trees but keep all trees" should be logically corrected.  
- Add missing key information: if position is unspecified, choose a reasonable area based on composition (near subject, empty space, center/edges).  

# Output Format Example
```json
{
   "Rewritten": "..."
}
'''

def polish_prompt(prompt, img):
    original_prompt = prompt
    prompt = f"{SYSTEM_PROMPT}\n\nUser Input: {prompt}\n\nRewritten Prompt:"
    success=False
    max_retries = 3
    retry_count = 0
    
    while not success and retry_count < max_retries:
        try:
            result = api(prompt, [img])
            # print(f"Result: {result}")
            # print(f"Polished Prompt: {polished_prompt}")
            if isinstance(result, str):
                result = result.replace('```json','')
                result = result.replace('```','')
                result = json.loads(result)
            else:
                result = json.loads(result)

            polished_prompt = result['Rewritten']
            polished_prompt = polished_prompt.strip()
            polished_prompt = polished_prompt.replace("\n", " ")
            success = True
        except Exception as e:
            print(f"[Warning] Error during API call (attempt {retry_count + 1}): {e}")
            retry_count += 1
            
    if not success:
        print(f"[Warning] Failed to polish prompt after {max_retries} attempts, using original prompt")
        return original_prompt
        
    return polished_prompt


def encode_image(pil_image):
    import io
    buffered = io.BytesIO()
    pil_image.save(buffered, format="PNG")
    return base64.b64encode(buffered.getvalue()).decode("utf-8")




def api(prompt, img_list, model="qwen-vl-max-latest", kwargs={}):
    import dashscope
    api_key = os.environ.get('DASH_API_KEY')
    if not api_key:
        raise EnvironmentError("DASH_API_KEY is not set")
    assert model in ["qwen-vl-max-latest"], f"Not implemented model {model}"
    sys_promot = "you are a helpful assistant, you should provide useful answers to users."
    messages = [
        {"role": "system", "content": sys_promot},
        {"role": "user", "content": []}]
    for img in img_list:
        messages[1]["content"].append(
            {"image": f"data:image/png;base64,{encode_image(img)}"})
    messages[1]["content"].append({"text": f"{prompt}"})

    response_format = kwargs.get('response_format', None)

    response = dashscope.MultiModalConversation.call(
        api_key=api_key,
        model=model, # For example, use qwen-plus here. You can change the model name as needed. Model list: https://help.aliyun.com/zh/model-studio/getting-started/models
        messages=messages,
        result_format='message',
        response_format=response_format,
        )

    if response.status_code == 200:
        return response.output.choices[0].message.content[0]['text']
    else:
        raise Exception(f'Failed to post: {response}')

# --- Model Loading ---
dtype = torch.bfloat16
device = "cuda" if torch.cuda.is_available() else "cpu"

# Load the model pipeline
pipe = QwenImageEditPipeline.from_pretrained("Qwen/Qwen-Image-Edit", torch_dtype=dtype).to(device)

# --- UI Constants and Helpers ---
MAX_SEED = np.iinfo(np.int32).max

# --- Main Inference Function (with hardcoded negative prompt) ---
@spaces.GPU(duration=180)
def infer(
    image,
    prompt,
    seed=0,
    randomize_seed=True,
    true_guidance_scale=1.0,
    num_inference_steps=50,
    rewrite_prompt=True,
    progress=gr.Progress(track_tqdm=True),
):
    """
    Generates an image using the local Qwen-Image diffusers pipeline.
    """
    # Hardcode the negative prompt as requested
    negative_prompt = " "
    
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    # Set up the generator for reproducibility
    generator = torch.Generator(device=device).manual_seed(seed)
    
    # Korean prompts will be automatically translated via polish_prompt() function
    
    print(f"Calling pipeline with prompt: '{prompt}'")
    print(f"Negative Prompt: '{negative_prompt}'")
    print(f"Seed: {seed}, Steps: {num_inference_steps}, Guidance: {true_guidance_scale}")
    try:
        if rewrite_prompt:
            prompt = polish_prompt(prompt, image)
            print(f"Rewritten Prompt: {prompt}")

        # Generate the image
        images = pipe(
            image,
            prompt=prompt,
            negative_prompt=negative_prompt,
            num_inference_steps=num_inference_steps,
            generator=generator,
            true_cfg_scale=true_guidance_scale,
            num_images_per_prompt=1
        ).images
        
        return images[0], seed
    
    except Exception as e:
        print(f"Error during inference: {e}")
        # Return the original image with error message
        return image, seed

# --- Examples and UI Layout ---
examples = []

css = """
#col-container {
    margin: 0 auto;
    max-width: 1024px;
}
#edit_text{
    margin-top: -62px !important
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.HTML('<h1 style="text-align: center; color: #6366f1; font-size: 3rem; font-weight: bold; margin: 2rem 0; font-family: system-ui, -apple-system, sans-serif; display: flex; align-items: center; justify-content: center; gap: 1rem;"><img src="https://huggingface.co/spaces/tchung1970/Qwen-Image-Edit/resolve/main/logo.png" alt="로고" style="height: 3rem; width: auto;"> 퀀 이미지 편집기</h1>')
        gr.Markdown("[더 알아보기](https://github.com/QwenLM/Qwen-Image)에서 Qwen-Image 시리즈에 대해 자세히 알아보세요. [Qwen Chat](https://chat.qwen.ai/)에서 체험하거나 [모델 다운로드](https://huggingface.co/Qwen/Qwen-Image-Edit)하여 ComfyUI나 diffusers로 로컬에서 실행해보세요.")
        with gr.Row():
            with gr.Column():
                input_image = gr.Image(
                    label="입력 이미지", 
                    show_label=False, 
                    type="pil",
                    height=400,
                    interactive=True,
                    placeholder="이미지를 여기로 끌어다 놓으세요"
                )

            result = gr.Image(
                label="결과", 
                show_label=False, 
                type="pil", 
                format="png",
                height=400,
                interactive=False
            )
        prompt = gr.Text(
                label="프롬프트",
                show_label=False,
                placeholder="편집 지시사항을 설명해주세요",
                container=False,
        )
        
        
        with gr.Row():
            run_button = gr.Button("편집!", variant="primary", size="lg", scale=1)
            with gr.Accordion("고급 설정", open=False):
                # Negative prompt UI element is removed here

                seed = gr.Slider(
                    label="시드",
                    minimum=0,
                    maximum=MAX_SEED,
                    step=1,
                    value=0,
                )

                randomize_seed = gr.Checkbox(label="시드 랜덤화", value=True)

                with gr.Row():

                    true_guidance_scale = gr.Slider(
                        label="가이던스 스케일",
                        minimum=1.0,
                        maximum=10.0,
                        step=0.1,
                        value=4.0
                    )

                    num_inference_steps = gr.Slider(
                        label="추론 단계 수",
                        minimum=1,
                        maximum=50,
                        step=1,
                        value=50,
                    )
                    
                    rewrite_prompt = gr.Checkbox(label="프롬프트 재작성", value=True)

        gr.Markdown("예시를 클릭하면 이미지와 한국어 프롬프트가 입력되며, 한국어 프롬프트는 '편집!' 버튼을 누를 때 자동으로 영어로 번역되어 AI가 처리합니다.")
        
        gr.Examples(
                label="예시",
                examples=[
                ["neon_sign.png", "텍스트를 'COOL NEON SIGN HERE'으로 변경해주세요"],
                ["cat_sitting.jpg", "고양이가 전통 한국 한복을 입고 있는 모습으로 만들어 주세요"],
                ["pie.png", "사진 스타일을 빈티지 만화책 스타일로 바꿔주세요"]],
                    inputs=[input_image, prompt], 
                    cache_examples=False,
                    examples_per_page=3)
        # Force update timestamp

    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=infer,
        inputs=[
            input_image,
            prompt,
            seed,
            randomize_seed,
            true_guidance_scale,
            num_inference_steps,
            rewrite_prompt,
        ],
        outputs=[result, seed],
    )

if __name__ == "__main__":
    demo.launch()