videogencn 🎬

中文文档

A Claude Code / OpenClaw skill for generating video clips with Chinese video models across four providers — Alibaba Bailian (Wan/PixVerse/Kling/Vidu/HappyHorse), Volcengine Ark (Jimeng/即梦), MiniMax (海螺 AI), and Tencent Hunyuan (混元).

Features

• Four modes, one script: prompt → t2v; --image → i2v; + --last-frame → kf2v; --ref name=img → r2v
• Seven model families across four providers: Bailian (Wan, PixVerse, Kling, Vidu, HappyHorse), Jimeng (Volcengine), MiniMax (Hailuo), Hunyuan (Tencent)
• Provider auto-detection: --provider flag or auto-detect from model name; backward compatible
• Local images just work: base64 for Wan/HappyHorse/Jimeng; auto-upload for PixVerse/Kling/Vidu/MiniMax
• Multi-shot narratives: wan2.7-t2v renders up to 15s with per-shot descriptions
• Audio control: --audio (third-party) / --no-audio (Wan)
• Resumable: long tasks print a task id; --task-id resumes polling

Pipeline

Install

In a coding agent — tell it:

help me to install https://github.com/Agents365-ai/videogenCN.git

365-Skills Marketplace (in Claude Code):

/plugin install videogencn@365-skills

Manual:

git clone https://github.com/Agents365-ai/videogenCN.git /tmp/videogencn
ln -s /tmp/videogencn/skills/videogencn ~/.claude/skills/videogencn  # global
# or: ln -s /tmp/videogencn/skills/videogencn .claude/skills         # project
# or: ln -s /tmp/videogencn/skills/videogencn ~/.openclaw/skills     # OpenClaw

Requirements

• Python 3.8+
• pip install requests
• At least one provider API key (see below)

Provider API Keys

Provider	Env Vars	Get Key At
Bailian (Wan/PixVerse/Kling/Vidu/HappyHorse)	DASHSCOPE_API_KEY	https://bailian.console.aliyun.com/
Jimeng (即梦)	ARK_API_KEY	https://console.volcengine.com/ark/
MiniMax (海螺 AI)	MINIMAX_API_KEY	https://platform.minimax.io
Hunyuan (混元)	HUNYUAN_API_KEY	https://console.cloud.tencent.com/hunyuan

# Bailian (required for default provider)
export DASHSCOPE_API_KEY='your-api-key'

# Jimeng (optional)
export ARK_API_KEY='your-api-key'

# MiniMax (optional)
export MINIMAX_API_KEY='your-api-key'

# Hunyuan (optional)
export HUNYUAN_API_KEY='your-api-key'

Optional environment variables:

Variable	Purpose
DASHSCOPE_API_BASE	Bailian region: cn (default) / sg / us or a full URL
DASHSCOPE_VIDEO_MODEL	Bailian default model override

Quick Start

Natural language (in Claude Code):

用万相生成一段 5 秒的视频:一只柴犬在樱花树下奔跑 用即梦生成一段 10 秒的竖屏视频:赛博朋克雨夜街头

Command line:

# Text-to-video (Bailian default)
python scripts/generate_video.py "A shiba inu running under cherry blossoms" out.mp4

# Image-to-video (animate a still)
python scripts/generate_video.py "Camera slowly zooms in" out.mp4 --image photo.png

# Vertical short-video clip
python scripts/generate_video.py "Cyberpunk rainy street" city.mp4 --ratio 9:16

# Jimeng (即梦)
python scripts/generate_video.py "城市日落延时摄影" sunset.mp4 --provider jimeng --duration 10

# MiniMax (海螺)
python scripts/generate_video.py "海浪拍打礁石" ocean.mp4 --provider minimax --duration 6

# Hunyuan (混元)
python scripts/generate_video.py "金黄色的麦田在秋风中起伏" field.mp4 --provider hunyuan --duration 5

Models

Bailian (百炼) — 5 families

Family	Models	Modes	Duration
Wan 通义万相	wan2.7-t2v-* (t2v default), wan2.6-i2v-flash (i2v default), wan2.5/2.2/wanx2.1 series	t2v, i2v	up to 15s
PixVerse 爱诗	pixverse/pixverse-{c1,v6,v5.6}-{t2v,it2v,kf2v,r2v}	all four	1–15s
Kling 可灵	kling/kling-v3-video-generation, kling/kling-v3-omni-video-generation	t2v, i2v, kf2v (+r2v on omni)	3–15s
Vidu	vidu/viduq3-{pro,turbo}_{text2video,img2video,start-end2video}, viduq2*	t2v, i2v, kf2v	q3: 1–16s
HappyHorse	happyhorse-{1.1,1.0}-{t2v,i2v}	t2v, i2v	3–15s

Jimeng (即梦 / Volcengine Ark)

Family	Models	Modes	Duration
Jimeng Seedance	doubao-seedance-2-0-260128 (default), doubao-seedance-2-0-fast-260128, doubao-seedance-1-5-pro-251215, doubao-seedance-1-0-pro	t2v, i2v	up to 15s

✅ Tested: t2v 5s, ~4 min, 5.6 MB MP4

MiniMax (海螺 AI)

Family	Models	Modes	Duration
MiniMax	video-01 (t2v/i2v default)	t2v, i2v	6s

✅ Tested: t2v 6s, ~3 min, 2.9 MB MP4

Hunyuan (混元 / Tencent)

Family	Models	Modes	Duration
Hunyuan	hy-video-1.5 (t2v/i2v default), yt-video-2.0 (i2v, experimental), yt-video-fx (i2v, experimental), yt-video-humanactor (i2v, experimental)	t2v, i2v	5–10s

⚠️ Note: --duration and --seed supported on hy-video-1.5 only. --resolution, --ratio, --audio, and --camera-motion are not yet available for Hunyuan. Experimental models (yt-video-*) are i2v-only with basic prompt+image support.

Run python scripts/generate_video.py --list-models for the current list. Third-party Bailian families are Beijing region (cn) only.

Cost: video generation bills per second of output (roughly $0.04–0.14/s depending on model and resolution). Result URLs expire after 24h — the script downloads immediately.

Support

If this project helps you, consider supporting the author:

Author

Agents365-ai

• Bilibili: https://space.bilibili.com/441831884
• GitHub: https://github.com/Agents365-ai

License

This project is licensed under CC BY-NC 4.0 — free for non-commercial use. Commercial use requires permission. See the LICENSE file for details.