Nodes/Symbiotica/NS WaveSpeed Wan 2.5 Text to Image
ComfyUI Node

NS WaveSpeed Wan 2.5 Text to Image

The only door to a model Alibaba never open-sourced

By symbiotica-ai·Created 6 months ago·Updated about 13 hours ago· 2
NS WaveSpeed Wan 2.5 Text to Image
  • client
  • output_image
prompt
size1024*1024
negative_prompt
enable_prompt_expansionfalse
seed-1
enable_sync_modefalse

Here's the uncomfortable fact that makes this node interesting: there is no Wan 2.5 to download. Alibaba stopped shipping open weights after Wan 2.2 - 2.5, 2.6, and 2.7 are all API-only (wan-video.md). So NSWaveSpeedWan25TextToImage from symbiotica-ai/comfyui-nodes isn't a convenience wrapper around a model you could run yourself - it's one of the only doors to that model, period. WaveSpeed was the first platform to host Wan 2.5, and this node is a ComfyUI-shaped key to it.

How it works

Prompt in, IMAGE out. The node POSTs to WaveSpeed (/api/v3/alibaba/wan-2.5/text-to-image) with your prompt, a size, an optional negative_prompt, a prompt-expansion toggle, and a seed, then downloads the resulting image back into ComfyUI as a tensor - so unlike the video nodes in this pack, the deliverable is a real IMAGE you can wire straight into a Save Image. enable_sync_mode defaults off (submit + poll), and Cancel works during the poll.

The size field is a plain text string - 1024*1024 by default, and the tooltip says each dimension should sit between 768 and 1440 pixels. That's a free-text field, so it accepts what you type; respect the range or the API will push back. enable_prompt_expansion (off by default) lets Wan rewrite and enrich your prompt before generating - handy when your prompt is a sentence rather than a paragraph, and worth turning off once you've found a phrasing that works so it stops second-guessing you.

The inputs that matter

  • prompt - the image description.
  • size - width*height, 768–1440 per dimension.
  • negative_prompt - what to avoid; Wan steers well with it.
  • enable_prompt_expansion - auto-enrich the prompt.
  • seed - -1 for random, or set it to reproduce.
  • client - the WAVESPEED_AI_API_CLIENT socket from the NS WaveSpeed Client node.

Output: output_image (IMAGE) - straight into a preview or Save Image.

Install and keys

# ComfyUI Manager: search "Symbiotica", or:
cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt

Restart, add the NS WaveSpeed Client node with your wavespeed.ai key (widget, Settings → Symbiotica, or WAVESPEED_API_KEY), wire client in. No model files.

Common issues

A malformed size gets rejected. It expects width*height - use the *, not x. And keep both numbers in the 768–1440 range.

You can't run this one locally, and that's not your fault. Every "just download the 2.5 weights" answer you'll find is wrong - the community hit this wall repeatedly; 2.2 is the newest Wan with open weights. If the cost or the cloud round-trip bugs you, the local fallback is Wan 2.2 on your own GPU, which is free after electricity but a step down in quality (wan-video.md). For 2.5 itself, pay-per-call is the whole deal (external-api-nodes.md).

Categoryneuralsins/WaveSpeed

Inputs (7)

NameTypeDefaultDescription
clientWAVESPEED_AI_API_CLIENT
promptSTRINGText description for image generation
sizeoptSTRING1024*1024Image resolution (width*height). Range: 768~1440 pixels per dimension
negative_promptoptSTRINGDescribe what you don't want in the image
enable_prompt_expansionoptBOOLEANfalseAutomatically expand and enhance the prompt
seedoptINT-1-1–2147483647Random seed for reproducible results. -1 for random seed
enable_sync_modeoptBOOLEANfalseWait for generation to complete before returning

Outputs (1)

NameTypeDescription
output_imageIMAGE