一句成语生成AI视频故事：Hermes Agent全流程实操

import requests

BASE_URL = "https://blazeai.boxu.dev/api/v1"
API_KEY="blz_xxxxx"  # Your Blaze AI key

# 文生图
response = requests.post(
    f"{BASE_URL}/images/generations",
    headers={"Authorization": f"Bearer ***              "model": "flux-dev",
            "prompt": "A golden cocoon opening under moonlight, butterfly emerging, cinematic 8K",
            "n": 1,
            "size": "1024x1024"
        }
    }
)
image_url = response.json()["data"][0]["url"]

# 图生视频
response = requests.post(
    f"{BASE_URL}/videos/generations",
    headers={"Authorization": f"Bearer ***              "model": "kling",
            "image_url": image_url,
            "duration": 5,
            "mode": "standard"
        }
    )
video_url = response.json()["url"]

import subprocess, json

prompt = "金色蚕茧在月光下缓缓裂开，一只发光蝴蝶破茧而出，数字粒子特效"
payload = json.dumps({
    "model": "agnes-image-2.1-flash",
    "prompt": prompt,
    "n": 1,
    "size": "768x1024"
})

# Windows 需用 subprocess curl 绕过 SSL 问题
with open("_payload.json", "w") as f:
    f.write(payload)

result = subprocess.run([
    "curl", "-sS", "-X", "POST",
    "https://apihub.agnes-ai.cn/v1/images/generations",
    "-H", "Authorization: Bearer ***    "-H", "Content-Type: application/json",
    "-d", "@_payload.json"
], capture_output=True, text=True, timeout=90)

data = json.loads(result.stdout)
image_url = data["data"][0]["url"]

#!/usr/bin/env python3
"""
成语AI视频生成器 - 完整流水线
输入：一个成语
输出：一段完整的AI视频故事
"""
import json, os, subprocess, time, requests

# ============ 配置 ============
BLAZE_BASE = "https://blazeai.boxu.dev/api/v1"
BLAZE_KEY = "blz_xxxxx"
AGNES_BASE = "https://apihub.agnes-ai.cn/v1"
AGNES_KEY = "sk_xxxxx"
IDIOM = "破茧成蝶"

def generate_script(idiom):
    """用 Hermes Agent 生成结构化故事脚本"""
    return {
        "title": f"{idiom}—— 一个关于蜕变的故事",
        "scenes": [
            {
                "id": 1,
                "description": "暗金色的茧悬挂在古老树枝上，月光透过树叶洒下",
                "narration": "在幽深的森林里，一只小小的茧正在等待命运的召唤。",
                "mood": "mysterious",
                "duration": 5
            },
            {
                "id": 2,
                "description": "茧开始微微颤动，金色裂纹从内部蔓延开来",
                "narration": "内心的力量开始涌动，那是生命对自由的渴望。",
                "mood": "tension",
                "duration": 5
            },
        ]
    }

def generate_image(prompt, outpath):
    """使用 Agnes 生成场景图"""
    payload = json.dumps({
        "model": "agnes-image-2.1-flash",
        "prompt": prompt, "n": 1, "size": "1024x1024"
    })
    pl = outpath.replace(".png", ".json")
    with open(pl, "w") as f:
        f.write(payload)
    result = subprocess.run([
        "curl", "-sS", "-X", "POST",
        f"{AGNES_BASE}/images/generations",
        "-H", f"Authorization: Bearer ***        "-H", "Content-Type: application/json",
        "-d", f"@{pl}"
    ], capture_output=True, text=True, timeout=90)
    os.remove(pl)
    data = json.loads(result.stdout)
    img_url = data["data"][0]["url"]
    subprocess.run(["curl", "-sS", "-o", outpath, img_url], timeout=60)
    return outpath

def generate_video(image_path, outpath, duration=5):
    """使用 BlazeAI 将静态图转为视频"""
    response = requests.post(
        f"{BLAZE_BASE}/videos/generations",
        headers={"Authorization": f"Bearer {BLAZE_KEY}"},
        json={
            "model": "kling-v2",
            "image_url": upload_image(image_path),
            "duration": duration,
            "mode": "standard"
        }
    )
    video_url = response.json()["url"]
    subprocess.run(["curl", "-sS", "-o", outpath, video_url], timeout=120)
    return outpath

def stitch_videos(video_paths, output_path):
    """使用 ffmpeg 拼接多段视频"""
    with open("concat_list.txt", "w") as f:
        for v in video_paths:
            f.write(f"file '{v}'\n")
    subprocess.run([
        "ffmpeg", "-f", "concat", "-safe", "0",
        "-i", "concat_list.txt",
        "-c:v", "libx264", "-crf", "18",
        "-preset", "medium",
        "-c:a", "aac", "-b:a", "192k",
        "-movflags", "+faststart",
        "-y", output_path
    ], check=True)

def main():
    script = generate_script(IDIOM)
    video_clips = []
    for scene in script["scenes"]:
        print(f"处理场景 {scene['id']}...")
        img_path = f"scene_{scene['id']}.png"
        generate_image(scene["description"], img_path)
        vid_path = f"scene_{scene['id']}.mp4"
        generate_video(img_path, vid_path, scene["duration"])
        video_clips.append(vid_path)
    stitch_videos(video_clips, f"{IDIOM}_final.mp4")
    print(f"完成！输出: {IDIOM}_final.mp4")

if __name__ == "__main__":
    main()

# 方法1: 直接拼接（同编码参数）
ffmpeg -f concat -i clips.txt -c copy output.mp4

# 方法2: 带过渡效果的拼接（推荐）
ffmpeg -i clip1.mp4 -i clip2.mp4 -filter_complex "
[0:v]trim=0:4,setpts=PTS-STARTPTS[v0];
[1:v]trim=0:4,setpts=PTS-STARTPTS[v1];
[v0][v1]xfade=transition=fade:duration=1:offset=3[outv];
[0:a]atrim=0:4,asetpts=PTS-STARTPTS[a0];
[1:a]atrim=0:4,asetpts=PTS-STARTPTS[a1];
[a0][a1]acrossfade=d=1[outa]
" -map "[outv]" -map "[outa]" output.mp4

# 将所有片段统一为相同参数，避免拼接失败
for f in scene_*.mp4; do
    ffmpeg -i "$f" \
           -vf "scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:(ow-iw)/2:(oh-ih)/2" \
           -r 30 -c:v libx264 -crf 18 -c:a aac -ar 48000 \
           "uniform_$f"
done

# 生成旁白（BlazeAI TTS）
curl -X POST "https://blazeai.boxu.dev/v1/audio/speech" \
  -H "Authorization: Bearer *** \
  -d '{"model":"elevenlabs","input":"在幽深的森林里...","voice":"zh_female"}' \
  -o narration.mp3

# 合并音视频
ffmpeg -i video.mp4 -i narration.mp3 -c:v copy -c:a aac output.mp4

# 烧字幕
ffmpeg -i output.mp4 -vf "subtitles=story.srt" -c:a copy final.mp4