wan / v2.7 / fal / createImage

Wan 2.7 is the newest closed Wan, and it's... fine, apparently. The community's loudest reaction was a one-line dunk ("Its shit tbh, compared to other closed source models") plus general indifference, though the one real evaluation found it genuinely strong on small-detail textures in close-ups and stills. You can't run it locally - no weights exist - so this node is how it gets into ComfyUI: a cloud recipe on the fal engine, billed in Buzz.

The distinctive thing about this node isn't the model, it's the use_pro toggle. That's the "pay for the better tier" switch - same node, same prompt, and use_pro on means the job runs on Civitai's higher-cost lane (more compute, presumably better results, definitely more Buzz). If you're cost-conscious, that's the single most important knob in the pack: it's the difference between a cheap experiment and a bill you notice.

Inputs and outputs

• prompt (required).
• use_pro (default false) - the pro-tier toggle. Off for drafts, on for the keeper.
• guidance_scale - 3.5 default.
• image_size - preset string (square_hd).
• enable_prompt_expansion, enable_safety_checker, loras, negative_prompt, seed, quantity (up to 12), output_format, image_metadata.

Outputs: images (IMAGE), errors, workflow_id, raw_json. The raw_json is worth wiring to a text viewer on this node specifically - it's where you'll see what the pro-tier job actually charged.

Install

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

Or ComfyUI Manager → "Civitai Comfy Nodes" → Install → restart. Authenticate (CIVITAI_API_TOKEN or sidebar) - required, since every run bills your account.

Gotchas

The Wan 2.7 line is where the "open Wan" story fully dead-ended, and this pack faithfully reflects it: no weights, no local fallback, and a use_pro gate on top. Before you spend Buzz, the honest advice from the community's own comparison: 2.7's best territory is static and slow-pan work with fine detail - faces, fabric, water. It showed "visible pose discontinuity" on heavy character motion, and reviewers kept pointing at free open models (LTX2 especially) as the better value. So the practical play: draft at use_pro off, evaluate the still against a free local model, and only spend the pro tier when the output genuinely needs it. The node works exactly as advertised; whether 2.7 is worth its price is a question the node's use_pro toggle forces you to answer for yourself, one job at a time.