Alibaba Wan 2.7 Video Extend (Official Edition)
Wan 2.7 video extend
- api_config
- video
- request_id
- urls
Every AI video model has a hard ceiling on length, and Wan 2.7 is no exception. Video extend is the workaround: take a clip you already generated and continue it, coherently, instead of starting a fresh shot that won't match anything. This node does exactly that - feed it footage, tell it (optionally) what should happen next, and it returns the clip plus its continuation. It runs on Alibaba's cloud through the BizyAirPlus pack, so there's no local GPU requirement and no model download.
It's the narrowest node in the Wan video family, and that's a feature: while VEdit rewrites what's already in your footage, this one just keeps the story going. The tooltips are minimal, which is its way of saying "attach a video, set a length, go."
How it works
The required inputs are resolution and duration (2–15 seconds, default 15 - the tooltip's plain "Video length in seconds" is the whole story). You attach the source clip, the node uploads it to the BizyAir cloud, Wan 2.7 continues the motion and scene, and you get back video, request_id, and urls. A continuation node's core promise is that the new section matches the old one in motion, lighting, and subject - that's what you're paying for here.
Everything else is optional. prompt describes the content you want the continuation to cover; negative_prompt says what you don't want to see; seed, watermark, and prompt_extend (LLM smart-overwrite for short prompts) round it out. skip_error lets the workflow carry on with a placeholder if the request fails, and api_config overrides the base URL/API key for a single request.
The inputs that matter
- duration - the target length for the extended output. This is where the cost lives, same as the rest of the Wan API family.
- prompt - optional but recommended. A continuation with no prompt has nothing to steer it; even a one-line "she keeps walking through the market" grounds the next few seconds.
Installation
The one-time pack install shared by all BizyAirPlus nodes. ComfyUI Manager, search BizyAirPlus, or:
cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt
Restart ComfyUI, add a bizyair.ai API key via the ON-button prompt, Settings > BizyAirPlus > API Key, or BIZYAIR_API_KEY. Nothing heavy downloads - the execution happens server-side.
Common issues
- Continuation doesn't match the source - a longer
durationgives the model more room to drift. Shorter extensions hold consistency better; chain two extends instead of one long one. - Unexpected length - remember
durationis the target output length in seconds, capped at 15, and it's what you're billed on. - Cloud failures - key validity and network first, then the ComfyUI console for the detailed error.
- Pack toggling ON unexpectedly - BizyAirPlus auto-installs and its action-bar button might be sitting ON, which can interfere with local-only runs. Set it to OFF when you want to work locally.
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| resolution | COMBO | 720P | Output video resolution |
| duration | INT | 152–15 | Video length in seconds |
| promptopt | STRING | Video generation prompts, which describe the desired video content to be generated | |
| negative_promptopt | STRING | Backtracking, describing content you don't want to see in a video | |
| seedopt | INT | 0-1–2147483647 | Random seed, -1 means automatically generated |
| watermarkopt | BOOLEAN | false | Whether to add a watermark identification |
| prompt_extendopt | BOOLEAN | false | If prompt smart overwrite is enabled, use the large model to overwrite the input prompt |
| api_configopt | BIZYAIR_OPENAPI_CONFIG | 单次请求覆盖base_url和api_key | |
| skip_erroropt | BOOLEAN | false | 开启后遇到错误不中断工作流,输出对应类型的错误占位符 |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| request_id | STRING | — |
| urls | STRING | — |