Shima Workflow Image Creator

Every workflow you share needs a thumbnail, and most people make them the sad way: screenshot the graph, or paste text over a random generated image in an editor. Shima Workflow Image Creator is the pack's answer - a node that composites a proper workflow cover - big title, subtitle, logo, drop shadows - as a single IMAGE output. It's a poster generator for your pipeline, and it's the most opinionated node in this batch.

What it makes and how

The node draws on a 1024×1024 black canvas (or any base_image you feed it), pastes an optional logo from your ComfyUI input folder, then centers your text with the font you pick. All done with PIL in the backend: LANCZOS-resized logo, a blurred drop shadow behind it (unless you use a "Center Large" position, which swaps to a flat dark or light background), and your main title plus subtitle rendered bottom-up with configurable spacing.

The font selection is genuinely nice - the pack ships seven bundled fonts including Noto Sans JP, Poppins, Cal Sans, and EB Garamond, so there's no font download to hunt down and the dropdown populates itself. One trick worth knowing: if you connect a shima.commonparams bundle (the pack's shared settings DICT) and leave use_commonparams on, the bundled model_type overrides text_main - so a cover can auto-label itself "SDXL" or "FLUX" from your pipeline settings instead of you typing it.

The inputs that matter

• text_main - the big title (default "SDXL")
• text_sub - the subtitle line (default "workflow")
• font_name - one of the 7 bundled fonts
• main_font_size / sub_font_size - sizes (10–500)
• subtext_position - subtitle Above or Below the title
• logo_file - dropdown of files in your ComfyUI input/ folder (ships with example.png)
• use_logo / logo_position - toggle and place the logo (corners, center, or "Center Large (Dark/Light BG)" variants)
• save_mode - write the result to disk, overwriting the previous run

Optional extras: base_image, custom_name (if set, it becomes the only text line and the subtitle is dropped), shima.commonparams, use_commonparams, allow_external_linking. Output is a single IMAGE - and the node is an output node, so it shows in the preview and the workflow queues to it.

About save_mode

This is the gotcha most people hit first. save_mode is off by default, and the tooltip is refreshingly direct: "Save to disk (overwriting previous run)." Off means the cover only goes to the preview - nothing lands in ComfyUI/output/. On means it writes <your main text>.png to the output folder, sanitized of characters Windows hates, and yes, it overwrites the same file every run. That's deliberate: it's meant to keep a single current cover, not accumulate cover_0001.png. If you want an actual SaveImage-style history, wire the output into ComfyUI's own Save Image node instead.

Installing it

Part of the Shima pack (KDB-USJP/shima_wf). ComfyUI Manager → search "Shima", or:

cd ComfyUI/custom_nodes
git clone https://github.com/KDB-USJP/shima_wf

Restart after. The pack auto-installs cg-use-everywhere and Impact-Pack on first launch; neither matters much for this node, but they'll be there anyway.

Where people get burned

The logo_file dropdown reads your ComfyUI input directory, so if you're on a fresh install there's no example.png - use the upload/refresh in the widget or drop your own logo file into ComfyUI/input/. Also note the docstring mentions auto-downloading Google Fonts, but the shipped code only reads the bundled fonts/ folder and silently falls back to PIL's default bitmap font if a file's missing - so if your text looks blocky, the font file didn't load, not your settings.

Is it essential? No. But if you share workflows on Civitai-style galleries or inside the Shima ecosystem, a decent cover costs you nothing and makes the difference between "workflow" and "product." For that job, it's the right tool.