Wan Image-to-Video Effect Generator

This is the party-trick node of the ru4ls/ComfyUI_Wan pack. Give it a photo, pick a template like flying, rose, or frenchkiss, and it animates the image with that preset effect - no prompt engineering, no prompt at all, actually. WanI2VEffectGenerator ("Wan Image-to-Video Effect Generator") is a cloud Wan 2.1 model that turns a still into a short effects-driven video clip.

Where every other node in this pack takes a prompt, this one takes a template - twenty canned effects (squish, rotation, poke, inflate, dissolve, carousel, singleheart, the dance1-3 trio, mermaid, graduation, dragon, money, flying, rose, crystalrose, hug, frenchkiss, coupleheart). You pick one, it runs. It's the closest thing to TikTok-ify-your-photo in this pack, and for wedding montages or silly group photos it's genuinely fun.

The catch you need to know first

This feature only works in the mainland China region. The region input defaults to mainland_china for a reason, and unlike the other nodes, the international region won't cut it. That means you need a separate API key for the China endpoint: DASHSCOPE_API_KEY_CHINA in config/.env, alongside your international key. People skip that line in the README, hit a 403 or an auth error, and blame the node. It's the key step.

How it works and what to set

Same async DashScope pattern as the rest of the pack: submit image_url + template to the video endpoint, poll every 10 seconds until the clip is done (up to 60 tries, so give it a few minutes), and download the mp4 to your output_dir. Your two outputs are video_file_path and video_url; wire the path into a "Load Video (Path)" node from ComfyUI-VideoHelperSuite to preview it.

The inputs are refreshingly few:

• image_url - required. Must be a publicly reachable URL; Alibaba's servers fetch it, so local files won't work.
• template - the effect, 20 options. squish, rotation, poke, inflate, dissolve, carousel are the playful ones; dance1/2/3, flying, and the heart trio are the romantic set.
• resolution - 480P, 720P (default), or 1080P. Higher is slower and costlier.
• seed, output_dir - the usual.

No negative prompt, no watermark toggle, no prompt_extend. The template is the prompt.

Install

ComfyUI Manager, search ComfyUI_Wan, or:

cd ComfyUI/custom_nodes
git clone https://github.com/ru4ls/ComfyUI_Wan.git
pip install -r ComfyUI_Wan/requirements.txt

Restart ComfyUI, then in the pack's config/ folder copy .env.template to .env and set both keys - DASHSCOPE_API_KEY for anything international and DASHSCOPE_API_KEY_CHINA for this node. Dependencies are minimal (requests, Pillow, numpy, python-dotenv) because the compute is remote and billed per job.

The honest take

It's a toy, but a well-made one - the effects are polished and you can batch a stack of portraits through it in an afternoon. Just don't expect the mainland_china region switch to be invisible. Between the separate API key and China-endpoint latency, this is the one node in the pack where setup friction actually shows up. Get the China key sorted first and the rest is genuinely one-click.