AI 做漫剧：从剧本到成片的工程化流水线与代码实现

pip install openai requests moviepy

import os, json, base64, requests
from openai import OpenAI
from moviepy.editor import ImageClip, AudioFileClip, concatenate_videoclips

client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

def gen_script(topic, episodes=3):
    """生成结构化漫剧剧本"""
    prompt = f"""写一个{topic}的漫剧剧本，共{episodes}集。
每集30秒，输出JSON：
{{"title":"", "episodes":[{{"no":1,"shots":[
{{"shot":1,"desc":"画面","line":"台词","duration":3}}]}}]}}"""
    resp = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": prompt}],
        response_format={"type": "json_object"},
        temperature=0.7,
    )
    return json.loads(resp.choices[0].message.content)

def gen_image(prompt, save_path):
    """调用 Stable Diffusion WebUI API 生成分镜图"""
    url = "http://127.0.0.1:7860/sdapi/v1/txt2img"
    payload = {
        "prompt": prompt + ", manga style, clean line, cinematic",
        "negative_prompt": "low quality, bad hands, text, watermark",
        "steps": 28, "width": 768, "height": 1024,
        "seed": 12345,  # 固定种子提升一致性
    }
    r = requests.post(url, json=payload).json()
    with open(save_path, "wb") as f:
        f.write(base64.b64decode(r["images"][0]))

def compose(shots, out="manju.mp4"):
    """按分镜合成视频：图片 + 音频 + 时长"""
    clips = []
    for img, audio in shots:
        a = AudioFileClip(audio)
        c = ImageClip(img).set_duration(a.duration).set_audio(a)
        clips.append(c)
    final = concatenate_videoclips(clips, method="compose")
    final.write_videofile(out, fps=24, codec="libx264", audio_codec="aac")

if __name__ == "__main__":
    script = gen_script("都市异能漫剧", episodes=1)
    print(json.dumps(script, ensure_ascii=False, indent=2))
    # 实际生产中：遍历 shots，调用 gen_image 生成图，
    # 调用 TTS 生成音频，最后传入 compose 合成。