Text Overlay

Need a label on an image - a character name, a test caption, a watermark - without opening a paint program? UC_TextOverlayNode draws text directly onto any image in your graph, with a background box, transparency, padding, and flexible positioning. It's the kind of utility you don't think about until a batch of comparison renders all need their parameters stamped on them.

What it is

A pure PIL-based compositor. It takes your image, renders your text at a chosen size and color, optionally draws a padded background bar behind it at a transparency you control, and composites the result. The output is an IMAGE you can preview, save, or feed onward - so you can put labels into a pipeline and still chain processing after.

How it works

Colors are hex strings (text_color, bg_color) - no # needed, though one is tolerated. The background is drawn as RGBA using bg_transparency from 0.0 (invisible) to 1.0 (fully opaque), so at 0.5 you get a classic semi-transparent caption bar. Font loading tries Arial first, then falls back to DejaVu Sans on Linux, then to PIL's default - so it runs everywhere, though the exact font look differs between Windows and Linux.

Positioning is the part worth understanding. top, bottom, left, right all default to -1, which means center on that axis. To pin text to a corner, set one offset on each axis (say top and right) and leave the other at -1. And if you'd rather place things relative to image size, flip use_percentage on and the same four fields become percentages 0-100. That's the knob you want when a fixed-pixel offset at 512x512 looks wrong at 2K.

The inputs that matter

• image - the canvas.
• text - what to draw (multiline, so captions can span lines).
• font_size - 32 default, up to 1024.
• text_color / bg_color - hex strings.
• draw_background - toggles the box behind the text; off means just the text.
• bg_padding / bg_transparency - the box's padding and opacity.
• use_percentage + top/bottom/left/right - positioning, absolute pixels or relative.

Output: a single image, ready for a Save node.

Installing it

Part of ComfyUI-UtilsCollection:

cd ComfyUI/custom_nodes
git clone https://github.com/silveroxides/ComfyUI-UtilsCollection

Restart ComfyUI (or use Manager, search "ComfyUI-UtilsCollection"). The pack's Python dependencies are light - opencv-python and typing-extensions - and this node needs no model downloads. It does require a recent ComfyUI build since the pack targets the current Core node API.

Common issues

Most people trip on placement, not rendering. Remember -1 means "center on this axis", so top=-1, right=-1 puts text top-right - and if you set both top and bottom, the tooltip notes one wins as the offset for the other, so pick one per axis. Second, hex colors: pass six or eight hex digits, not color names. And on headless Linux boxes, the fallback font renders differently than your Windows screenshot - if you care about the exact typeface, that's the one thing this node can't control.