Load Image With Text 💎 Just Nodes

Load Image With Text is the reading half of the round-trip that Save Image With Text writes. Standard ComfyUI's Load Image gives you an image and a mask; this node gives you an image, a mask, and whatever text was embedded in the file. If the PNG was saved by the pack's Save node, the caption you attached comes right back out on the text output.

That makes it the natural front end for any image-to-X workflow built on your own dataset. Feed it images you generated with captions, and the caption lands on a string output ready to feed a text encoder, a prompt-conditioning setup, or a classification chain. The image goes to the image output, an alpha mask (if the file has one) goes to mask, and the metadata text goes to text - three outputs, one node, no chaining.

The image input is a dropdown of every image file in your input directory (with the upload button, so you can drag a file in and it lands there). Same image handling as core ComfyUI: EXIF rotation is applied, RGB conversion, and the mask is derived from the alpha channel - inverted to ComfyUI's convention where 1.0 means "erase this," exactly like the built-in Load Image. The node even hashes the file for its IS_CHANGED, so it re-runs only when the file actually changes rather than every queue.

The one thing to internalize: it reads PNG just_text metadata, not sidecar .txt files. Save Image With Text embeds text in PNGs but writes separate .txt files for JPEG and WebP. This node reads the embedded metadata only - so if you saved a JPEG with the pack, Load Image With Text will happily load the picture but return an empty text. For text round-tripping, stay on PNG.

Installing it

Same as the whole pack: ComfyUI Manager → search "Just Nodes" → install → restart, or cd ComfyUI/custom_nodes && git clone https://github.com/Arroz-11/ComfyUI-Just-Nodes.git. No models, no extra dependencies, MIT licensed.

Gotchas

• Text comes from PNG just_text metadata only. JPEG/WebP sidecar .txt files are not read by this node.
• The mask is optional - files without an alpha channel get a dummy 64×64 zero mask, so don't build logic that assumes a real mask exists.
• Images must live in the input directory (or be uploaded through the picker) - this is ComfyUI's Load Image convention, not an arbitrary folder like Image From Folder.