图片｜GPT-Image-2-C 旧版修图

This is a compatibility node, and the README/schema say so plainly: it "restores the original gpt-image-2-all edit entry" with the model renamed gpt-image-2-c. Translation: when Tikpan retired the old gpt-image-2-all model, people's saved workflows broke - so the pack shipped TikpanGptImage2EditNode as the drop-in replacement so old graphs load and run again. The node itself is a clean image-edit call: base image in, instruction, optional mask and product reference, edited image out. If you're on a fresh workflow you can use the V2 official editor instead; if you're resurrecting an old one, this is the node that un-breaks it.

How it works

The required 底图 (base image) goes in along with 修改指令, and the relay's edit model re-renders the image. Optional 遮罩_Mask limits the edit: white areas are redrawn, black areas are preserved - classic masked inpainting semantics. The optional 产品参考图 feeds a product/style reference into the result. Outputs are 重绘结果图 (IMAGE) and 渲染日志 (STRING).

The inputs that actually matter

• API_密钥 - sk- from tikpan.com.
• 底图 - the image to edit. Required, IMAGE type.
• 修改指令 - what to change (default: "change the background to a beach, keep the subject consistent").
• 输出尺寸 - 沿用底图尺寸 (follow the base image, default) or 512/1K/2K/4K.
• 画面比例 - 沿用底图比例 or 1:1 / 16:9 / 9:16 / 21:9 / 4:3 / 3:4.
• 品质 - standard (faster) or hd (better detail, default).

Optional: 遮罩_Mask, 产品参考图, and the 中转站地址 (defaults to https://tikpan.com - leave it).

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/htrert/ComfyUI-Tikpan-Pro

restart (or Manager → "Tikpan"). No models, no heavy deps.

Where people get burned

• It's the legacy lane, not the future. Tikpan explicitly frames this as the old entry with the model renamed. New work should use TikpanGptImage2OfficialEditV2 (multipart /v1/images/edits) - this node exists so old graphs don't die. If you're building fresh, don't reach for the legacy path first.
• Mask polarity. White = redraw, black = keep. Invert it and you'll preserve the thing you wanted to change.
• Model drift. gpt-image-2-c is the renamed replacement - if the upstream ever renames again, the node's 模型 field is a single-choice enum, so an old workflow may silently need the node updated to follow.
• Cost still applies. It's a compatibility node, not a free one - edits meter like everything else in the pack.

The right way to use it: keep it for graph-recovery and for any old tutorial workflow that references gpt-image-2-all, and know that the V2 editor is where the pack is putting its effort now. Compatibility nodes are the glue that keeps the ecosystem's history loadable - that's a genuinely useful job, just not the frontier.