#!/usr/bin/env python3
"""通过 API易 调用 Wan2.7 / HappyHorse 生成视频（文生 / 图生 / 参考图生 / 视频编辑）。纯标准库，零依赖。"""
import argparse
import base64
import json
import os
import shutil
import sys
import time
import urllib.error
import urllib.request
# 输出重定向到文件/管道时也逐行落盘，方便 Agent 后台跟踪进度
sys.stdout.reconfigure(line_buffering=True)
# DashScope 透传端点。绝不要用 /v1/videos 扁平路径——它会丢弃 media 字段
CREATE_URL = "https://api.apiyi.com/wan/api/v1/services/aigc/video-generation/video-synthesis"
TASK_URL = "https://api.apiyi.com/v1/tasks/{}"
# 家族 × 模式 → 模型 ID。编辑版命名不规则（happyhorse 是 1.0 且带连字符），别手拼
FAMILY_MODELS = {
"wan": {"t2v": "wan2.7-t2v", "i2v": "wan2.7-i2v",
"r2v": "wan2.7-r2v", "edit": "wan2.7-videoedit"},
"happyhorse": {"t2v": "happyhorse-1.1-t2v", "i2v": "happyhorse-1.1-i2v",
"r2v": "happyhorse-1.1-r2v", "edit": "happyhorse-1.0-video-edit"},
}
# r2v 参考素材上限：wan 图+视频合计 5，happyhorse 仅图、最多 9
MAX_REFS = {"wan": 5, "happyhorse": 9}
RATIOS = ("16:9", "9:16", "1:1", "4:3", "3:4")
# 出片是异步任务：720P/5s 实测约 70-140 秒，1080P/长视频可能 5 分钟以上
POLL_FIRST_DELAY = 15
POLL_INTERVAL = 8
POLL_TIMEOUT = 20 * 60
def load_api_key():
"""优先读环境变量；否则在脚本所在目录及其父目录找 .env。"""
key = os.environ.get("APIYI_API_KEY")
if key:
return key
here = os.path.dirname(os.path.abspath(__file__))
for d in (here, os.path.dirname(here)):
env_path = os.path.join(d, ".env")
if os.path.exists(env_path):
with open(env_path, encoding="utf-8") as f:
for line in f:
line = line.strip()
if line.startswith("APIYI_API_KEY") and "=" in line:
return line.split("=", 1)[1].strip().strip('"').strip("'")
return None
def project_root():
"""从脚本位置向上找包含 .git 或 .claude 的目录，作为项目根目录；找不到则用当前工作目录。"""
d = os.path.dirname(os.path.abspath(__file__))
while True:
if os.path.isdir(os.path.join(d, ".git")) or os.path.isdir(os.path.join(d, ".claude")):
return d
parent = os.path.dirname(d)
if parent == d:
return os.getcwd()
d = parent
def resolve_path(out):
"""纯文件名 → 存到 <项目根>/wan-output/ 下，确保好找；带目录成分则按给定路径。"""
if os.path.dirname(out):
return os.path.abspath(out)
out_dir = os.path.join(project_root(), "wan-output")
os.makedirs(out_dir, exist_ok=True)
return os.path.join(out_dir, out)
def media_source(src):
"""素材入参：URL / data: 原样透传，本地文件转 base64 data URI（实测两系列可用）。"""
if src.startswith(("http://", "https://", "data:")):
return src
if not os.path.exists(src):
sys.exit(f"素材文件不存在：{src}")
ext = src.lower().rsplit(".", 1)[-1]
mime = {"png": "image/png", "webp": "image/webp", "mp4": "video/mp4",
"mov": "video/quicktime"}.get(ext, "image/jpeg")
with open(src, "rb") as f:
return f"data:{mime};base64,{base64.b64encode(f.read()).decode()}"
def api_request(url, api_key, body=None, extra_headers=None):
"""网关会标 content-encoding: gzip 但实际未压缩，必须 Accept-Encoding: identity。"""
headers = {
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
"Accept-Encoding": "identity",
}
if extra_headers:
headers.update(extra_headers)
data = json.dumps(body).encode() if body is not None else None
req = urllib.request.Request(url, data=data, headers=headers,
method="POST" if body is not None else "GET")
try:
with urllib.request.urlopen(req, timeout=60) as r:
return json.loads(r.read())
except urllib.error.HTTPError as e:
raise RuntimeError(f"请求失败 HTTP {e.code}：{e.read().decode(errors='replace')[:800]}")
def download(url, path):
"""下载结果视频：OSS 签名直链，绝不能带 Authorization 头（带了 403）。"""
req = urllib.request.Request(url)
with urllib.request.urlopen(req, timeout=300) as r, open(path, "wb") as f:
shutil.copyfileobj(r, f)
return os.path.getsize(path)
def detect_mode(args):
if args.video:
return "edit"
if args.image:
return "i2v"
if args.ref_image or args.ref_video:
return "r2v"
return "t2v"
def build_media(args, mode):
media = []
if mode == "i2v":
media.append({"type": "first_frame", "url": media_source(args.image)})
elif mode == "r2v":
for src in args.ref_image:
media.append({"type": "reference_image", "url": media_source(src)})
for src in args.ref_video:
media.append({"type": "reference_video", "url": media_source(src)})
elif mode == "edit":
media.append({"type": "video", "url": media_source(args.video)})
for src in args.ref_image:
media.append({"type": "reference_image", "url": media_source(src)})
return media
def main():
api_key = load_api_key()
if not api_key:
sys.exit("未找到 API Key：请在技能目录的 .env 写一行 APIYI_API_KEY=sk-xxx"
"（令牌须勾选 Wan&HappyHorse 分组、计费模式为按量）")
parser = argparse.ArgumentParser(description="Wan2.7 / HappyHorse 出视频")
parser.add_argument("prompt", help="提示词（画面 + 运镜 + 氛围；多素材时用「图1/视频1」指代）")
parser.add_argument("--model", default="wan", choices=sorted(FAMILY_MODELS),
help="wan=Wan2.7（默认，更便宜）/ happyhorse=HappyHorse-1.1（画质取向）")
parser.add_argument("--resolution", default="720P", type=str.upper,
choices=("720P", "1080P"), help="分辨率，默认 720P（注意无 480P）")
parser.add_argument("--ratio", default=None, choices=RATIOS,
help="宽高比（传首帧图时忽略；happyhorse 文档未列此参数，仅显式传时发送）")
parser.add_argument("--duration", type=int, default=5,
help="时长 2-15 整数秒，默认 5；含参考视频时上限 10；编辑模式跟随源视频")
parser.add_argument("--negative", help="负向提示词（不想出现的内容，500 字符内）")
parser.add_argument("--no-prompt-extend", action="store_true",
help="关闭提示词智能扩写（默认开启，短提示词效果更好）")
parser.add_argument("--seed", type=int, default=None, help="随机种子，复现用")
parser.add_argument("-i", "--image", help="首帧图（本地路径或 URL，图生视频）")
parser.add_argument("--ref-image", action="append", default=[],
help="参考图，可重复（wan 与参考视频合计最多 5、happyhorse 最多 9）")
parser.add_argument("--ref-video", action="append", default=[],
help="参考视频 URL，可重复（仅 wan 支持）")
parser.add_argument("--video", help="待编辑视频 URL（视频编辑模式，需配 --ref-image）")
parser.add_argument("-o", "--out", default="output.mp4", help="输出文件名")
args = parser.parse_args()
if args.image and (args.ref_image or args.ref_video):
sys.exit("首帧模式（-i）与参考模式（--ref-image/--ref-video）互斥，一次只能用一种。")
if args.video and args.image:
sys.exit("视频编辑模式（--video）与首帧模式（-i）互斥。")
if args.video and not args.ref_image:
sys.exit("视频编辑模式需要至少 1 张参考图（--ref-image）。")
if args.model == "happyhorse" and args.ref_video:
sys.exit("happyhorse 不支持参考视频（--ref-video），仅 wan 支持。")
n_refs = len(args.ref_image) + len(args.ref_video)
if n_refs > MAX_REFS[args.model]:
sys.exit(f"{args.model} 参考素材最多 {MAX_REFS[args.model]} 个，当前 {n_refs} 个。")
if not 2 <= args.duration <= 15:
sys.exit("时长只支持 2-15 整数秒。")
if args.ref_video and args.duration > 10:
sys.exit("含参考视频时时长上限 10 秒。")
mode = detect_mode(args)
model = FAMILY_MODELS[args.model][mode]
input_part = {"prompt": args.prompt}
if args.negative:
input_part["negative_prompt"] = args.negative
media = build_media(args, mode)
if media:
input_part["media"] = media
parameters = {
"resolution": args.resolution,
"duration": args.duration,
"prompt_extend": not args.no_prompt_extend,
}
if args.ratio:
parameters["ratio"] = args.ratio
if args.seed is not None:
parameters["seed"] = args.seed
body = {"model": model, "input": input_part, "parameters": parameters}
try:
resp = api_request(CREATE_URL, api_key, body,
extra_headers={"X-DashScope-Async": "enable"})
except (RuntimeError, OSError) as e:
sys.exit(f"提交失败：{e}")
task_id = (resp.get("output") or {}).get("task_id") or resp.get("task_id")
if not task_id:
sys.exit(f"提交失败，响应：{json.dumps(resp, ensure_ascii=False)[:500]}")
print(f"任务已提交 model={model} task_id={task_id}，出片通常需要 2-5 分钟，开始轮询…")
t0 = time.time()
time.sleep(POLL_FIRST_DELAY)
while True:
try:
task = api_request(TASK_URL.format(task_id), api_key)
except (RuntimeError, OSError) as e: # 网络抖动不中断轮询
print(f" 轮询异常（继续）：{e}")
time.sleep(POLL_INTERVAL)
continue
status = str(task.get("status", "unknown")).lower()
progress = task.get("progress", "")
elapsed = round(time.time() - t0)
# progress 常停在 30（上游只有 0/10/30/100 几档），不代表卡住
print(f" [{elapsed:>4}s] status={status}" + (f" progress={progress}" if progress != "" else ""))
if status in ("completed", "failed"):
break
if time.time() - t0 > POLL_TIMEOUT:
sys.exit(f"轮询超时（{POLL_TIMEOUT}s）。任务仍在服务端，可稍后手动查询：\n"
f" GET {TASK_URL.format(task_id)}")
time.sleep(POLL_INTERVAL)
if status != "completed":
err = task.get("error") or task.get("fail_reason") or task
sys.exit(f"生成失败（status={status}）：{json.dumps(err, ensure_ascii=False)[:500]}"
"\n（失败任务不计费）")
result_url = task.get("result_url")
if not result_url:
sys.exit(f"任务完成但未返回视频地址：{json.dumps(task, ensure_ascii=False)[:500]}")
path = resolve_path(args.out)
size = download(result_url, path)
print(f"视频已保存至 {path}（{size / 1e6:.1f} MB，耗时 {round(time.time() - t0)}s，"
f"模型 {model}）")
if __name__ == "__main__":
main()
