ZML_简易_保存图像

Some save nodes are a full control panel. ZML_SaveImage has a dozen knobs: text blocks, counters, metadata stripping, sidecar txt files, size limits. And then there's this one. ZML_SimpleSaveImage keeps exactly three options - operation mode, save path, and the text block field - and the author's changelog describes it as "very tidy." That's the pitch. When you just want to save an image with your prompt inside it and no essay of settings in the way, this is the node.

It's the same PNG-with-embedded-prompt idea as its big brother: type your prompt into 文本块存储 and it gets written into the PNG as a comfy_text_block text chunk (compressed), readable later by ZML's load nodes. The metadata machinery underneath - filename safety, unique filenames, the "don't save 1x1 placeholders" guard, the #-# filename separator - is shared with the main save node, so behavior matches even though the surface doesn't.

The three knobs

• 图像 (required input) - the IMAGE tensor to save.
• 操作模式 - 保存图像 or 仅预览图像. Preview writes to temp so you can check a frame without committing it to the output folder.
• 保存路径 - defaults to output/ZML/%Y-%m-%d, strftime-aware, and it's the same path convention the main node uses. Leave it alone and you get dated subfolders for free.
• 文本块存储 - the prompt (or any text) to embed. Empty = a normal PNG save.

There's no output shown in the schema - it's a terminal node; the "output" is the file on disk and the preview in the UI.

Why it exists

Honestly, the value is subtractive. The main save node's defaults are already good - timestamped filenames, dated folders, unique-file collision handling. What Simple adds is that there's nothing left to fiddle with, which matters when you're on run 47 of a batch and the last thing you want is to think about whether the counter is reset and whether metadata is being stripped. The author built it for exactly that mood, and it shares the lifetime "总次数" counter file with ZML_SaveImage, so the Help-style total keeps counting across both nodes if you swap between them.

Where it bites: because there's no 存储格式 dropdown, you get PNG only - which is fine, because PNG is the only format that can carry the text block anyway. And there's no metadata-strip option here, so every saved file keeps the full ComfyUI workflow graph embedded. If your priority is disk space or not leaking your prompt everywhere, use ZML_SaveImage with 清除元数据 instead. If your priority is "just save it," this is the one.

Install

Part of the ComfyUI-ZML-Image pack: ComfyUI Manager → search "ComfyUI-ZML-Image", or cd ComfyUI/custom_nodes && git clone https://github.com/zml-w/ComfyUI-ZML-Image, then restart. No model downloads for this node; the heavy requirements in the pack's requirements.txt (ultralytics, opencv-python) belong to its YOLO and image-processing nodes. The UI is Chinese by default, so pick up the English patch when you install.