HT Save Image Plus

The stock Save Image node will get your PNG onto disk, and nothing more. HT Save Image Plus (from the HommageTools pack) is for the workflows where that isn't enough: you want a JPEG at a specific quality, a TIFF with compression, a mask saved as the alpha channel, a caption text file dropped next to every image, or a filter that skips the garbage outputs before they ever hit disk. If you've ever built a training dataset or a batch pipeline, you know exactly why that's useful.

What it does

It's an output node (category HommageTools/output), so nothing wires out of it - you put it at the end of the line and it writes files. The essentials:

• output_format - PNG, JPEG, or TIFF. Each has its own controls: jpeg_quality (1–100), tiff_compression (adobe_deflate or none). TIFF support is why the pack pulls in tifffile.
• mask - feed a mask in and it becomes the image's alpha channel (toggle with save_alpha). Great for cutouts and layered work.
• text_content - a text sidecar file saved alongside the image with the same base name (text_extension, text_encoding). This is the training-dataset trick: prompt → .txt file per image, ready for a caption loader.
• min_pixel_threshold - skip any image with fewer pixels than this. Perfect for batch jobs where half the outputs are tiny garbage.
• crop_to_alpha - crop the saved image to the alpha channel's content before writing.
• filename_prefix and add_sequence_number - prefix plus auto-incrementing numbers, or a fixed filename if you turn sequencing off.

The captioning feature

The headline extra is automatic captioning via Florence2 (enable_captioning). It's not bundled with this pack - you connect a florence2_model (FL2MODEL type) loaded by a separate Florence2 node, then pick a caption_type (caption / detailed_caption / more_detailed_caption) and tune generation with max_new_tokens, num_beams, do_sample, and seed. You can prepend/append fixed text with caption_pre and caption_post, and keep_model_loaded stops it from reloading the model every run. This is a real time-saver when you're captioning a dataset before training, but it's the one feature with a moving part: no Florence2 node installed, no captions, no error - the checkbox just does nothing.

Installing it

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/ArtHommage/HommageTools.git
cd HommageTools
pip install -r requirements.txt

Restart ComfyUI, or use ComfyUI Manager → "HommageTools". Remember the pack's own disclaimer: it's labeled perpetual alpha, not for production. For saving images that's a low-risk stance to take, but don't be shocked if an update renames a parameter.

Common issues

JPEG has no alpha. Feed it a mask with output_format set to JPEG and the transparency is gone; use PNG or TIFF when alpha matters. Metadata in JPEG/TIFF is also less reliable than PNG - save_metadata is really a PNG feature. And check output_dir: it defaults to your ComfyUI output folder, and the node will happily write anywhere you point it, which is convenient for dataset builds but easy to misplace files if you've got several pipelines running. Oh, and if min_pixel_threshold is set higher than your smallest legitimate output, you'll wonder where images went - it skips silently by design.