qianwen-video-generation

The saved excerpt is a snapshot from review. The external source remains the complete and most current version.

---
name: qianwen-video-generation
description: "Generate videos using Wan and HappyHorse models. Supports text-to-video, image-to-video, first+last frame, reference-based role-play, and video editing (VACE). TRIGGER when: user wants to create, generate, or edit video content, mentions video generation/animation/video clips/Wan/HappyHorse models, or explicitly invokes this skill by name (e.g. use qianwen-video-generation). DO NOT TRIGGER when: user wants to generate images (use qianwen-image-generation), understand/analyze existing videos (use qianwen-vision), text-only tasks."
compatibility: "Requires Python 3.9+ and curl. Cursor: auto-loaded. Claude Code: read this skill's SKILL.md before first use."
---

# Qwen Video Generation

Generate videos using Wan and HappyHorse models. All tasks are **asynchronous** — submit, then poll until
completion.
This skill is part of **QianWen-AI/qianwen-ai**.

> **⚠️ Critical Parameter Differences:**
> - **kf2v (First+Last Frame)**: Duration is **fixed at 5 seconds** — other values will fail. Output is **silent only**.
> - **Resolution parameters vary by model family, not mode alone**: Check the current model catalog before choosing `size`, `resolution`, or `ratio`. For example, `happyhorse-1.1-t2v` uses `resolution` + `ratio`, while `wan2.6-t2v` uses `size`.

## Skill directory

Use this skill's internal files to execute and learn. Load reference files on demand when the default path fails or you need details.

| Location | Purpose |
|----------|---------|
| `scripts/video.py` | Default execution — mode auto-detect, submit, poll, download |
| `references/execution-guide.md` | Fallback: curl for all 5 modes, code generation |
| `references/request-fields.md` | Field tables and audio handling by mode |
| `references/workflows.md` | Duration extensions, multi-shot, VACE pipelines |
| `references/polling-guide.md` | Polling patterns and timing |
| `references/merge-media.md` | Concat, trim, audio overlay — ffmpeg/moviepy recipes |
| `references/prompt-guide.md` | Per-mode prompt formulas, sound description, multi-shot structure |
| `references/examples.md` | Full script examples per mode |
| `references/sources.md` | Official documentation URLs |

## Security

**NEVER output any API key or credential in plaintext.** Always use variable references (`$DASHSCOPE_API_KEY` in shell, `os.environ["DASHSCOPE_API_KEY"]` in Python). Any check or detection of credentials must be **non-plaintext**: report only status (e.g. "set" / "not set", "valid" / "invalid"), never the value. Never display contents of `.env` or config files that may contain secrets.

**When the API key is not configured, NEVER ask the user to provide it directly.** Instead, help create a `.env` file with a placeholder (`DASHSCOPE_API_KEY=sk-your-key-here`) and instruct the user to replace it with their actual key from the [QianWen Console](https://platform.qianwenai.com/home/api-keys). Only write the actual key value if the user explicitly requests it.

## Key Compatibility

Both PAYG (`sk-ws-...`; legacy `sk-...`) and Token Plan (`sk-sp-...`) keys are supported. Detect
the API key type without exposing the Key:

```bash
python3 -c "
import sys; sys.path.insert(0, 'scripts')
from qianwen_lib import detect_api_key_type
print(detect_api_key_type('scripts/qianwen_lib.py'))
"
```

| Output | Meaning |
|--------|---------|
| `token-plan` | Token Plan key — use only models from the Token Plan catalog below. |
| `payg` | Pay-as-you-go key — full model catalog available. |
| `not-set` | No key configured. |

For Token Plan, fetch and read the current [Token Plan model catalog](https://alioth.alicdn.com/skills-info/models/references/qianwen-token-plan-models.md), then use an exact listed model. If CDN access fails, use the [local fallback](cdn/references/qianwen-token-plan-models.md).

Token Plan does not support local file upload; for i2v/r2v/kf2v modes, provide reference images/videos
as accessible URLs (`https://` or `oss://`) rather than l