给 AI 助手构建视觉传达能力 - Hooper 的博客

# scripts/generate.py 核心抽象

def submit_task(provider_cfg, prompt, size=None, urls=None):
    """提交生图任务，返回 task_id"""
    headers = {
        "Authorization": provider_cfg["api_key"],
        "Content-Type": "application/json",
    }
    payload = {"prompt": prompt}
    if size:
        payload["size"] = size
    if urls:
        payload["urls"] = urls if isinstance(urls, list) else urls.split(",")

    resp = requests.post(
        provider_cfg["submit_url"],
        headers=headers, json=payload, timeout=30,
    )
    resp.raise_for_status()
    data = resp.json()

    if data.get("code") != 200:
        raise Exception(f"提交失败: {data.get('msg', '未知错误')}")
    return data["data"]["id"]

def poll_result(provider_cfg, task_id):
    """轮询任务结果"""
    elapsed = 0
    while elapsed < provider_cfg.get("max_wait", 300):
        time.sleep(provider_cfg.get("poll_interval", 3))
        elapsed += interval

        resp = requests.get(
            provider_cfg["query_url"],
            params={"key": provider_cfg["api_key"], "id": task_id},
            timeout=15,
        )
        result = resp.json().get("data", {})
        status = result.get("status", 0)

        if status == 2:   # 完成
            return result["result"][0]
        elif status == 3: # 失败
            raise Exception(f"生成失败: {result.get('message', '')}")
        # status == 0 继续等待

# scripts/config.yaml
providers:
  gpt-image-2:
    api_key: "YOUR_KEY"
    submit_url: "https://api.wuyinkeji.com/api/async/image_gpt"
    query_url: "https://api.wuyinkeji.com/api/async/detail"
    default_size: "auto"
    poll_interval: 3
    max_wait: 300

A photorealistic portrait of an Asian woman around 25 years old.
Sharp, delicate features. Intense intelligent eyes with a mischievous
glint and a slightly arrogant look. Confident smirk, not a gentle smile.
Mid-length wavy dark brown hair with natural sheen. Warm morning
sunlight, candid photography, film-like color grading...

A photorealistic portrait of an Asian woman around 25 years old. She has delicate, non-influencer facial features with a slightly sharp, intelligent look, but her smile is sweet and warm. Her eyes are bright and clever with a mischievous, witty glint — she looks like someone quick-witted and sharp-tongued. Medium-length wavy dark brown hair with a natural shine, casually styled but not messy. She wears a simple white t-shirt or knit sweater with casually rolled-up sleeves. A minimalist ring or simple watch on her wrist. Expression: slight upward curve at the corners of her mouth, eyes carrying a playful, teasing look as if she's about to say something sarcastic. Setting: morning sunlight streaming through a window, in a cozy café or at a desk, authentic lifestyle atmosphere, not overly polished or studio-like. Warm natural lighting, soft shadows, film-like color grading, shallow depth of field, candid photography style.

Based on the reference image, keep the exact same pose, composition, clothing (elegant white satin dress, black pearl necklace), prop (large dark rose held near shoulder), and deep black studio background with soft warm portrait lighting. Transform the subject into a 25-year-old Asian woman. She has sharp, delicate features with intense, intelligent eyes that hold a mischievous glint. Her expression is a confident, slightly teasing smirk, not a gentle smile — witty and sharp-edged, as if she's about to say something clever and sarcastic. Her eyes say "I'm smarter than you and I know it" with a playful, loving teasing vibe. Mid-length wavy dark brown hair with natural sheen. Maintain the elegant, slightly dark and mysterious mood, cinematic photography style, realistic skin texture, high quality.