DP Save Image V2

ComfyUI's built-in Save Image node is fine for casual use and genuinely annoying the moment you need anything beyond it - it writes PNG, names files with a counter, and that's about it. DP Save Image V2 covers the gaps: a real filename you control, a format choice beyond PNG, a folder you name yourself, and - the feature that makes this worth knowing about specifically - an optional caption file saved alongside the image.

How it works

folder_name and file_name give you actual control over where things land and what they're called, instead of ComfyUI's default counter-based naming. image_format lets you save PNG, JPG, JPEG, or WEBP directly from this one node, rather than needing a separate converter node for anything but PNG. mode toggles between SAVE_IMAGE (writes to disk) and PREVIEW_ONLY (shows the result without saving) - handy for flipping a save node to a scratch preview while you're iterating, without deleting and re-adding a node.

The part worth calling out specifically: save_caption. Turn it on and, alongside the image file, the node writes a text caption using whatever you feed into prompt_and_caption_text (and optionally negative_or_other). That's the exact image-plus-matching-text-file pairing that LoRA training tools expect - captioning by hand or with a captioner and keeping a .txt next to each image is still the standard workflow for building a training dataset. If you're generating synthetic training images and want them export-ready for a LoRA trainer without a separate captioning pass afterward, this is the node doing double duty.

Inputs and outputs

Required:

• mode - SAVE_IMAGE (default) or PREVIEW_ONLY.
• image - what you're saving.
• folder_name (default dp_image_folder) / file_name (default image) - where it goes and what it's called.
• extra_text (default empty) - appended to the filename.
• add_size_to_name (boolean, default off) - append the image dimensions to the filename.
• save_caption (boolean, default off) - write a matching text caption alongside the image.
• image_format - PNG (default), JPG, JPEG, or WEBP.

Optional:

• prompt_and_caption_text / negative_or_other (default empty) - the caption content, used when save_caption is on.
• format_override (default empty) - override the chosen format on a per-run basis.

No data outputs - this is a terminal node (is_output_node: true), it writes to disk and that's the effect.

Installing it

ComfyUI Manager: search ComfyUI-Desert-Pixel-Nodes, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/DesertPixelAi/ComfyUI-Desert-Pixel-Nodes

Restart ComfyUI - no extra models, just image I/O.

Where people get tripped up

PREVIEW_ONLY is easy to leave on by accident after a debugging session and then wonder why a batch you ran produced no files - check mode first if your output folder is emptier than expected.

add_size_to_name and extra_text both modify the filename, and if you're relying on file_name alone to be exactly reproducible run to run (for matching against an external caption list, say), stacking those options in will change the actual filename on disk - know what's turned on before you build automation around exact filenames.

And if you turn on save_caption expecting a caption to just appear, remember it's a passthrough, not a generator - this node writes whatever text you feed into prompt_and_caption_text, it doesn't invent a caption from the image. You still need something upstream producing that text, whether that's your own prompt string or a captioning model's output.