Wan2.7 Image

Wan 2.7 is the newest numbered Wan, and like its 2.5 and 2.6 siblings it exists only behind an API - no open weights, no local checkpoint. This node is how you use it from ComfyUI. The "Image" suffix and the required width/height inputs make the story clear: this is the image-generation tier of Wan 2.7, a text-to-image node with a couple of genuinely interesting switches that the older API nodes don't have.

The interesting bits are settings.thinking and settings.sequential. Thinking mode is the model spending extra inference time reasoning about the generation before denoising - the tooltip says it improves quality at the cost of time, and it defaults to on. Sequential mode enables "image set output mode for generating multiple related images," which is how you coax a coherent series (same character, varied scenes) out of a single node instead of wrangling seeds yourself.

What you set

positivePrompt, width, and height are all required. Width/height run 768–2048 in steps of 16, so you can't request a 1×1 icon - the model has a floor.

• settings.thinking - on by default. Leave it on for quality; flip off for fast drafts. This is the main cost lever.
• settings.sequential - image-set mode for related outputs. Combine with numberResults to get a batch of connected images.
• referenceImages - optional IMAGE input for conditioning.
• seed - exposed and reproducible, so a fixed seed + settings gives you repeatable output.
• numberResults - up to 20 per run (each billed).
• advanced_json - merges JSON for settings.colorPalette and settings.editRegions - the two power features without dedicated widgets. colorPalette pins a palette; editRegions is where region-based edits go.

Output is image (IMAGE) → Preview/Save, with outputFormat JPG/PNG/WEBP and outputQuality.

Installing

Part of the Runware/ComfyUI-Runware pack.

ComfyUI Manager: search Runware, install, restart.

Manual:

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

Deps are runware-sdk, pillow, soundfile; no weights to fetch since there are none public. Runware API key from runware.ai/api-keys (Settings panel or RUNWARE_API_KEY).

Where people get burned

settings.thinking on by default means every run is the expensive path until you say otherwise. For iteration, switch it off; turn it back on for the final pass. And don't confuse "sequential" with free consistency - it generates a set of related images, not a guaranteed storyboard, and it multiplies the bill. Also note the 768px floor: if your workflow assumes you can generate small and upscale, this node won't cooperate.