SF WaveSpeed Wan 2.5 Image to Video Fast

Sometimes you don't need the flagship render, you need ten attempts. That's exactly what this node is for: Wan 2.5 image-to-video on WaveSpeed's fast endpoint. It's the iteration sibling of the pack's standard Image to Video node - same API family, faster turnaround, narrower options, and presumably a friendlier bill at the end of the month.

Wan 2.5 itself is API-only (Alibaba stopped shipping open weights after 2.2), so any ComfyUI path to it is a remote call. This node makes that call to /api/v3/alibaba/wan-2.5/image-to-video-fast, polls until the clip is done, and hands you a video_url. There's no local model anywhere in the chain - the only thing your machine does is hold the API key and wait.

How it works

Same as every node in this pack: an SF WaveSpeed Client wired into client, a prompt, and at run time the node POSTs your image and settings to the fast endpoint and polls. You get back one video_url string. Nothing downloads, nothing runs locally, no VRAM involved.

The inputs that matter

• image - the still to animate, as a URL string.
• prompt - the motion description. Keep it focused; a fast endpoint isn't where you do your best prompt engineering, but it's where you test which kind of prompt works before spending on the slow tier.
• resolution - 720p or 1080p. Note the difference from the standard node, which also offers 480p: Fast only goes down to 720p. That's the trade for the speed.
• audio - an optional audio URL (wav/mp3, 3–30s, ≤15MB) that guides generation, if you want audio-synced motion.
• duration - 5 or 10 seconds.
• enable_prompt_expansion - off by default; toggles the API's automatic prompt enhancement.
• negative_prompt, seed, enable_sync_mode - the usual trio, behaving exactly as their tooltips say.

Output: video_url, a string.

Installing it

ComfyUI Manager, search comfyui-stillfront, install. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/razvanmatei-sf/comfyui-stillfront.git
cd comfyui-stillfront
pip install -r requirements.txt

You need a WaveSpeed API key from wavespeed.ai: config.ini (copy config.ini.tmp first), the WAVESPEED_API_KEY env var, or the SF WaveSpeed Client's api_key widget all work. Ignore the README's stale stillfront/... clone URL - the repo is razvanmatei-sf/comfyui-stillfront.

Where people get burned

The image input is a URL string, not an IMAGE tensor - a stock Load Image node won't connect. You need an upload-style node that emits a URL. This is the pack-wide convention and the most common first stumble.

Also, "fast" here is a WaveSpeed endpoint choice, not a local speed-LoRA hack - don't expect the kind of quality tricks people apply to open Wan 2.2. It's a deliberate quality/speed trade on the hosted model. When you find a prompt and framing that work at Fast, that's the signal to re-run the winner through the standard node for the full quality render.