Bubba Add Text Overlay (Metadata)

Bubba Add Text Overlay (Metadata) stamps the generation details directly onto the image - the checkpoint name, a freeform info line, and the positive and negative prompts, each toggleable and positioned. Where Bubba Save Image puts that data inside the file where nobody can see it, this node paints it on the picture so everyone can. That's the point of the overlay: the "workflow included" culture from the KB's ecosystem essay meets people who consume images in a Discord feed, where embedded metadata is invisible.

It's also genuinely useful for your own iterations. Stuck trying to remember which of six nearly-identical renders used which checkpoint? Paint the model name and the prompt onto the image and the comparison becomes self-documenting. Compare sheets, batch reviews, sharing proofs with someone who doesn't open files in editors - this is the node for all of it.

How it works

It reads the BUBBA_METADATA from a pipe or a metadata socket (or you can skip those and wire an image directly, in which case it overlays whatever's in default metadata). Four independent toggles control what gets painted:

• show_model - the checkpoint/model name.
• show_info - the info line.
• show_positive - the positive prompt.
• show_negative - the negative prompt.

Each has its own position dropdown (top or bottom), so you can stack the model name up top and the full prompts along the bottom. background_color (default #000000AA, a translucent black) sets the text backing, font_size scales it, and overlay_mode - a boolean, on by default - is what makes the text overlay actually render over the image; leave it on unless you're staging.

The outputs are the overlaid image plus the original pipe and metadata passed through, so it slots into a Bubba pipeline without breaking it - overlay, then save with Bubba Save Image.

The inputs that matter

For a first run just flip on show_model and show_positive, pick a position, and leave the rest. font_size (default 40) is the knob you'll actually fiddle with - long prompts need smaller text. Everything else is set-and-forget.

How to install it

Standard Bubba_Nodes pack install: ComfyUI Manager → "Bubba Nodes" → install → restart, or

cd ComfyUI/custom_nodes
git clone https://github.com/bubbafett5611/bubba_nodes.git

restart, pip install -r requirements.txt if needed. Requires ComfyUI v0.27.0+ (V3 node API). No model downloads.

Gotchas

Two things to watch. First, empty metadata is invisible - this node overlays what's in the metadata, and if your workflow never populated it (e.g. you didn't use a Bubba model loader that records the model name), the show_model line is blank even though the toggle is on. Run a Bubba checkpoint loader or the model comparison loader upstream to populate it. Second, and this is the big one from the KB's metadata essay: overlaying is burned in. The text becomes part of the pixels, so it won't strip cleanly and it destroys the "clean image" as a master. Keep an un-overlaid save downstream if you might want the clean version later - overlay into a preview branch, save the raw image on the main branch. Also note that long negative prompts at 40px will push the text block off-canvas on smaller images; drop font_size or disable show_negative for small outputs. The pack is young with no community discussion yet, so the README's metadata notes are the best reference on which fields the overlay can draw from.