wan / v2.7 / fal / editImage

Editing is where the Wan line's still-image reputation was built, and this is the newest iteration of it. Wan 2.7 edit takes one to four reference images plus a prompt and produces an edited result - think "change this character's outfit while keeping the pose," "put this prop into that scene." The multi-image input is the interesting part: with up to four references you're doing composition-style edits, not just img2img relighting, which is closer to what people actually mean by "editing" in 2026.

No open weights, no local run - like its text-to-image sibling, this node is the only door into Wan 2.7 editing from a Comfy graph, running on the fal engine behind Civitai's Orchestration API for Buzz.

Inputs and outputs

• prompt (required) - the edit instruction.

• images (required, and the tooltip is unusually blunt): "Reference images for editing (1-4 images required)." Note that unlike the 2.5 i2i node, here it's in the required block - the node refuses to run without it. Feed it via a Load Image output or by chaining other image nodes.

• use_pro - default false; the pro-tier toggle, same as the createImage sibling. On for the keeper, off for drafts.

• guidance_scale - 3.5. Lower = trust the prompt's edit more; higher = keep closer to the references.

• image_size, enable_prompt_expansion, enable_safety_checker, loras, negative_prompt, seed, quantity, output_format, image_metadata.

Outputs: images (IMAGE), errors, workflow_id, raw_json.

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).

Gotchas

Three things to keep in mind. First, the 1–4 image requirement means the node fails fast if you forget to wire images - that's the pack's early-preview rigor showing up as a local error instead of a confusing cloud 400, which is genuinely good behavior. Second, every reference image leaves your machine and gets processed on Civitai's servers under their moderation - for multi-image edit work that's a real privacy decision, and there's no local 2.7 to fall back on.

Third, and this is the one that decides whether the node is worth it: the community read on 2.7's editing strengths is "cleanest small-detail textures on close-ups" with pose continuity problems on heavy motion - but a static multi-image edit is exactly where that profile is strong. If your edit job is "rework this character/prop set in place," 2.7 edit is plausibly the best use of your Buzz in the whole Wan family. If it's "dramatically move/relight people," expect the same pose wobble the reviewers flagged. Pick your jobs accordingly and you'll be glad the node exists.