ConfigurableDrawText

Sometimes the final image needs words on it - a caption, a watermark, a label - and you'd rather not fire up a paint program for every output. ConfigurableDrawText renders text onto an image using a text style defined elsewhere in the graph. It's the "draw" half of the Flux Continuum pack's text-overlay duo; DrawTextConfig is the "configure" half, and the two are meant to be used together.

How it works

The node takes your text, an image, and a TEXT_STYLE object, and renders the text over the image using Pillow. It builds a transparent overlay canvas, draws each line with the configured font and color, composites on optional shadow and background, then alpha-composites the whole thing over your image and returns the result as an IMAGE.

Details worth knowing: multi-line text splits on newlines, the text box is positioned by the style's alignment and offsets, and a direction of "rtl" reverses each line (for right-to-left scripts). The node operates on the first image in a batch - the source processes IMAGE[0] - so if you feed a batch, you get text on the first frame and unchanged pixels for the rest. That's a gotcha if you batch captions.

Inputs and outputs

• TEXT - STRING, multiline. The text to draw; newlines become line breaks.
• TEXT_STYLE - the style object from DrawTextConfig (font, size, color, alignment, shadows, offsets).
• IMAGE - the image to render onto.

Output is one IMAGE, the composite result.

Installing it

Bundled with robertvoy/ComfyUI-Flux-Continuum:

cd ComfyUI/custom_nodes
git clone https://github.com/robertvoy/ComfyUI-Flux-Continuum

or ComfyUI Manager ("Flux Continuum"), restart. The pack ships two bundled fonts (Inter and ShareTechMono) in its fonts/ directory, so there's nothing to download - the font choice comes from DrawTextConfig's dropdown.

Troubleshooting

Text not showing? Check the style's color isn't #00000000 (transparent) - the default background is transparent black, and if you accidentally set the text color to something transparent, you get nothing. Text off-canvas? The alignment + offset math can push the text box outside the image; drop the offsets to 0 and pick center alignment to sanity-check. And remember: single-image processing means a batch of images only gets text on the first one. If you need per-image captions, you want a node that loops the batch.