zhuchaowe/qwen_agent
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

增加seedream

Browse Source
This commit is contained in:
朱潮 2026-03-26 16:00:07 +08:00
parent 79206aafea
commit 40f7c55a9e
3 changed files with 353 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

105
skills/seedream/SKILL.md Executable file
View File

@ -0,0 +1,105 @@
---
name: seedream
description: 使用火山引擎 Seedream/Seedance API 生成高质量图片和视频。适用于文生图、图生图、文生视频、图生视频以及生成关联组图的场景。
---
# Seedream
本 Skill 封装了火山引擎（Volcengine）的 Seedream-4.5 图片生成和 Seedance-1.5 视频生成能力，支持文生图、图生图、文生视频、图生视频。
## 使用方法
### 文生图
生成单张图片（默认分辨率）：
```bash
python {baseDir}/scripts/generate_image.py --prompt "一只赛博朋克风格的猫"
```
指定分辨率（如 2K, 4K 或具体像素）：
```bash
python {baseDir}/scripts/generate_image.py --prompt "壮丽的山川日出" --size "2K"
```
### 图生图
提供参考图片 URL：
```bash
python {baseDir}/scripts/generate_image.py --prompt "将其风格变为印象派" --image "https://example.com/input.jpg"
```
### 生成组图
生成一组内容关联的图片（最多 15 张）：
```bash
python {baseDir}/scripts/generate_image.py --prompt "一个宇航员在不同行星上的探险经历" --sequential --max-images 4
```
## 参数说明
- `--prompt`: (必选) 图像生成的文本描述。
- `--model`: (可选) 模型 ID，默认为 `doubao-seedream-4-5-251128`。
- `--size`: (可选) 图像尺寸。支持 `2K`, `4K` 或 `2048x2048` 等格式。
- `--image`: (可选) 参考图 URL 或 Base64 编码。
- `--sequential`: (可选) 开启组图生成功能。
- `--max-images`: (可选) 组图生成的最大图片数量（1-15）。
## 工作流
1. 调用 `generate_image.py` 脚本。
2. 脚本会输出以 `MEDIA_URL: ` 开头的图片链接。
3. 提取链接并使用 Markdown 语法展示：`![Generated Image](URL)`。
4. 除非用户要求，否则无需下载图片。
## 注意事项
- Seedream-4.5 支持中英文提示词。
- 组图功能仅在 Seedream-4.5/4.0 模型中有效。
- 确保提供的图片 URL 可公开访问。
---
## 视频生成
### 文生视频
```bash
python {baseDir}/scripts/generate_video.py --prompt "无人机以极快速度穿越复杂障碍或自然奇观，带来沉浸式飞行体验"
```
### 图生视频
提供参考图片 URL：
```bash
python {baseDir}/scripts/generate_video.py --prompt "让画面中的人物动起来" --image "https://example.com/input.jpg"
```
指定视频时长（5 或 10 秒）：
```bash
python {baseDir}/scripts/generate_video.py --prompt "日出延时摄影" --duration 10
```
固定摄像机位：
```bash
python {baseDir}/scripts/generate_video.py --prompt "一只猫在桌上跳跃" --camera-fixed
```
### 视频参数说明
- `--prompt`: (必选) 视频生成的文本描述。
- `--model`: (可选) 模型 ID，默认为 `doubao-seedance-1-5-pro-251215`。
- `--image`: (可选) 参考图 URL，用于图生视频。
- `--duration`: (可选) 视频时长，5 或 10 秒，默认 5。
- `--camera-fixed`: (可选) 固定摄像机位。
- `--no-watermark`: (可选) 去除水印。
- `--poll-interval`: (可选) 轮询间隔秒数，默认 5。
- `--max-wait`: (可选) 最大等待秒数，默认 600。
### 视频生成工作流
1. 调用 `generate_video.py` 脚本。
2. 脚本会自动创建异步任务并轮询结果。
3. 完成后输出以 `MEDIA_URL: ` 开头的视频链接。
4. 提取链接并展示给用户（视频链接可直接访问）。
5. 视频生成通常需要 1-3 分钟，请耐心等待。
### 视频注意事项
- Seedance 视频生成是异步任务，脚本会自动轮询等待结果。
- 支持中英文提示词。
- 图生视频时，确保图片 URL 可公开访问。

87
skills/seedream/scripts/generate_image.py Executable file
View File

@ -0,0 +1,87 @@
#!/usr/bin/env python3
# /// script
# requires-python = ">=3.10"
# dependencies = [
# "requests>=2.31.0",
# ]
# ///
import argparse
import json
import os
import sys
import requests
def generate_image(prompt, model, size, api_key, image_input=None, sequential=False, max_images=1):
url = "https://ark.cn-beijing.volces.com/api/v3/images/generations"
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {api_key}"
}
payload = {
"model": model,
"prompt": prompt,
"size": size,
}
if image_input:
payload["image"] = image_input
if sequential:
payload["sequential_image_generation"] = "auto"
payload["sequential_image_generation_options"] = {"max_images": max_images}
else:
payload["sequential_image_generation"] = "disabled"
try:
response = requests.post(url, headers=headers, json=payload)
response.raise_for_status()
result = response.json()
if "data" in result and len(result["data"]) > 0:
for item in result["data"]:
if "url" in item:
print(f"MEDIA_URL: {item['url']}")
elif "b64_json" in item:
# In a real scenario, we might want to save this to a file,
# but for this skill we follow the pattern of providing a URL if possible.
print("ERROR: Received base64 data but expected URL.")
else:
print(f"ERROR: No image data in response. Full response: {json.dumps(result)}")
except requests.exceptions.RequestException as e:
print(f"ERROR: API request failed: {e}")
if hasattr(e, 'response') and e.response is not None:
print(f"Response body: {e.response.text}")
sys.exit(1)
def main():
parser = argparse.ArgumentParser(description="Generate images using Volcengine Seedream API.")
parser.add_argument("--prompt", required=True, help="Text prompt for image generation")
parser.add_argument("--model", default="doubao-seedream-4-5-251128", help="Model ID or Endpoint ID")
parser.add_argument("--size", default="2048x2048", help="Image size (e.g., 2K, 4K, 2048x2048)")
parser.add_argument("--api-key", help="Volcengine API Key")
parser.add_argument("--image", help="Input image URL or base64 (optional)")
parser.add_argument("--sequential", action="store_true", help="Enable sequential image generation (group)")
parser.add_argument("--max-images", type=int, default=1, help="Max images for sequential generation (1-15)")
args = parser.parse_args()
api_key = args.api_key or os.environ.get("VOLC_API_KEY")
if not api_key:
print("ERROR: API key is required. Provide via --api-key or VOLC_API_KEY environment variable.")
sys.exit(1)
generate_image(
prompt=args.prompt,
model=args.model,
size=args.size,
api_key=api_key,
image_input=args.image,
sequential=args.sequential,
max_images=args.max_images
)
if __name__ == "__main__":
main()

161
skills/seedream/scripts/generate_video.py Executable file
View File

@ -0,0 +1,161 @@
#!/usr/bin/env python3
# /// script
# requires-python = ">=3.10"
# dependencies = [
# "requests>=2.31.0",
# ]
# ///
import argparse
import json
import os
import sys
import time
import requests
API_BASE = "https://ark.cn-beijing.volces.com/api/v3/contents/generations/tasks"
def create_video_task(prompt, model, api_key, image_url=None, duration=5, camera_fixed=False, watermark=True):
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {api_key}"
}
content = [
{
"type": "text",
"text": f"{prompt} --duration {duration} --camerafixed {str(camera_fixed).lower()} --watermark {str(watermark).lower()}"
}
]
if image_url:
content.append({
"type": "image_url",
"image_url": {"url": image_url}
})
payload = {
"model": model,
"content": content
}
try:
response = requests.post(API_BASE, headers=headers, json=payload)
response.raise_for_status()
result = response.json()
task_id = result.get("id")
if not task_id:
print(f"ERROR: No task id in response. Full response: {json.dumps(result)}")
sys.exit(1)
return task_id
except requests.exceptions.RequestException as e:
print(f"ERROR: API request failed: {e}")
if hasattr(e, 'response') and e.response is not None:
print(f"Response body: {e.response.text}")
sys.exit(1)
def poll_video_task(task_id, api_key, poll_interval=5, max_wait=600):
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {api_key}"
}
url = f"{API_BASE}/{task_id}"
elapsed = 0
while elapsed < max_wait:
try:
response = requests.get(url, headers=headers)
response.raise_for_status()
result = response.json()
status = result.get("status")
if status == "succeeded":
content = result.get("content", {})
# content can be a dict with "video_url" key or a list of items
video_url = ""
if isinstance(content, dict):
video_url = content.get("video_url", "")
elif isinstance(content, list):
for item in content:
if isinstance(item, dict) and item.get("type") == "video_url":
video_url = item.get("video_url", {}).get("url", "")
break
elif isinstance(item, str) and item.startswith("http"):
video_url = item
break
if video_url:
print(f"MEDIA_URL: {video_url}")
return
# Fallback: try to find any URL in the response
resp_str = json.dumps(result)
import re
urls = re.findall(r'https?://[^\s"]+\.(mp4|mov|avi|webm)[^\s"]*', resp_str)
if urls:
print(f"MEDIA_URL: {urls[0]}")
return
print(f"ERROR: Task succeeded but no video URL found. Full response: {json.dumps(result)}")
return
if status == "failed":
error = result.get("error", {})
print(f"ERROR: Task failed. Code: {error.get('code')}, Message: {error.get('message')}")
sys.exit(1)
# Still running, wait and retry
print(f"STATUS: {status} (elapsed {elapsed}s, polling every {poll_interval}s...)", file=sys.stderr)
time.sleep(poll_interval)
elapsed += poll_interval
except requests.exceptions.RequestException as e:
print(f"ERROR: Polling failed: {e}")
if hasattr(e, 'response') and e.response is not None:
print(f"Response body: {e.response.text}")
sys.exit(1)
print(f"ERROR: Timed out after {max_wait}s waiting for task {task_id}")
sys.exit(1)
def main():
parser = argparse.ArgumentParser(description="Generate videos using Volcengine Seedance API.")
parser.add_argument("--prompt", required=True, help="Text prompt for video generation")
parser.add_argument("--model", default="doubao-seedance-1-5-pro-251215", help="Model ID")
parser.add_argument("--api-key", help="Volcengine API Key")
parser.add_argument("--image", help="Input image URL for image-to-video")
parser.add_argument("--duration", type=int, default=5, choices=[5, 10], help="Video duration in seconds (5 or 10)")
parser.add_argument("--camera-fixed", action="store_true", help="Fix camera position")
parser.add_argument("--no-watermark", action="store_true", help="Disable watermark")
parser.add_argument("--poll-interval", type=int, default=5, help="Polling interval in seconds")
parser.add_argument("--max-wait", type=int, default=600, help="Max wait time in seconds")
args = parser.parse_args()
api_key = args.api_key or os.environ.get("VOLC_API_KEY")
if not api_key:
print("ERROR: API key is required. Provide via --api-key or VOLC_API_KEY environment variable.")
sys.exit(1)
print(f"STATUS: Creating video task...", file=sys.stderr)
task_id = create_video_task(
prompt=args.prompt,
model=args.model,
api_key=api_key,
image_url=args.image,
duration=args.duration,
camera_fixed=args.camera_fixed,
watermark=not args.no_watermark
)
print(f"STATUS: Task created: {task_id}, polling for result...", file=sys.stderr)
poll_video_task(
task_id=task_id,
api_key=api_key,
poll_interval=args.poll_interval,
max_wait=args.max_wait
)
if __name__ == "__main__":
main()