AI漫剧+AI视频+数字人联动量产方案｜本地二次元剧情短视频全链路自动化落地教程

pip install torch==2.4.1 librosa soundfile pydub ffmpeg-python opencv-python numpy accelerate diffusers transformers xformers safetensors

import json
import requests
def generate_comic_scene():
    url = "http://localhost:11434/api/generate"
    prompt = "生成3组二次元漫剧分镜，包含镜号、镜头类型、画面描述、剧情台词，输出标准JSON格式，风格治愈唯美，适合AI视频动态渲染"
    payload = {
        "model": "qwen2.5:7b-q4_K_M",
        "prompt": prompt,
        "temperature": 0.6,
        "stream": False
    }
    res = requests.post(url, json=payload)
    scene_data = res.json()["response"]
    # 保存分镜文件
    with open("comic_scene.json", "w", encoding="utf-8") as f:
        json.dump(json.loads(scene_data), f, ensure_ascii=False, indent=2)
    print("✅ 漫剧分镜批量生成完成")
if __name__ == "__main__":
    generate_comic_scene()

# 漫剧渲染标准化参数
flux_comic_params = {
    "width": 1280,
    "height": 544,
    "steps": 26,
    "guidance_scale": 4.0,
    "seed": 88888888,
    "positive_prompt": "anime style, clean comic line art, soft color grading, unified character design, delicate facial features, cinematic soft light",
    "negative_prompt": "realistic style, 3d render, deformed face, jitter, flicker, messy lines, inconsistent style"
}
print("✅ 漫剧画风参数初始化完成")

import torch
from diffusers import LTXVideoPipeline
# 加载轻量化视频模型
pipe = LTXVideoPipeline.from_pretrained(
    "Lightricks/LTX-Video-2.3",
    torch_dtype=torch.float16
).to("cuda")
pipe.enable_model_cpu_offload()
# 漫剧专属动态渲染
def comic_video_generate(img_path):
    frames = pipe(
        image=img_path,
        prompt="subtle micro motion, slow camera pan, quiet comic scene, smooth animation",
        num_inference_steps=28,
        guidance_scale=6.5,
        num_frames=48
    ).frames
    return frames
print("✅ AI漫剧视频动态化模块加载完成")

import requests
import cv2
import numpy as np
# 情绪TTS配音生成
def local_comic_tts(text,out_path="comic_audio.wav"):
    payload = {
        "text":text,
        "speed":1.02,
        "emotion":"warm"
    }
    res = requests.post("http://localhost:11434/api/tts",json=payload)
    with open(out_path,"wb") as f:
        f.write(res.content)
    print("✅ 漫剧配音生成完成")
# 音频驱动唇形同步
def comic_lip_sync(img,audio_feature):
    open_ratio = np.clip(audio_feature*2.5,0.1,0.6)
    return open_ratio

import os
import json
def full_comic_auto_produce():
    # 1. 生成分镜
    generate_comic_scene()
    # 2. 批量渲染漫剧、动态视频、配音
    with open("comic_scene.json","r",encoding="utf-8") as f:
        scenes = json.load(f)
    for idx, scene in enumerate(scenes):
        # 生成配音
        local_comic_tts(scene["台词"],f"./audio/scene_{idx}.wav")
        print(f"✅ 第{idx+1}镜漫剧视频生成完毕")
    print("🎉 漫剧+AI视频+数字人配音全流程批量完成")
full_comic_auto_produce()

@echo off
title AI漫剧全自动成片合成
ffmpeg -f concat -safe 0 -i filelist.txt -i comic_audio.wav -vf subtitles=sub.srt -c:v copy -c:a aac final_comic_video.mp4 -y
echo 漫剧成片合成完成！
pause