BD Bulk Save

Here's the workflow that used to be ugly: a PBR derivation spits out albedo, normal, roughness, metallic, AO, ORM, ARM - seven images that all need saving with proper names. The old way was seven parallel save nodes, each with its own filename logic, each one an opportunity for a typo and a graph so wide it needs two monitors. BD Bulk Save kills that. It takes up to 16 inputs and writes them all in a single run, naming each file from a one-line-per-slot label list. The node description literally tells you the intended use: "Wire BD_DerivePBR's 7 outputs to input_1..input_7, set labels accordingly, and all 7 files save in a single Run. No queueing."

The inputs that matter

• input_1 … input_16 - the data to save. They're typed * (ANY), so images, masks, strings, whatever. Empty slots are skipped - you can wire three inputs today and nine tomorrow without touching anything else.
• labels - one label per line, aligned with the wired inputs after skipping empty slots. albedo on line one names input_1's file. If your label list is shorter than the wired inputs, the extra files save with no suffix.
• label_prefix - what gets prepended to each label to form the file suffix. Default is _, so albedo becomes _albedo and the file ends ..._albedo_001.png. Set it to "" if you want no separator.
• context_id - for the pack's Save Context system. Empty + exactly one BD Save Context registered = auto-pick. Otherwise specify which context's path template to resolve against.
• custom_vars - extra key=value lines layered on top of the context's variables, so one batch can dump into subfolder=PBR while another uses subfolder=textures.
• format - PNG by default, JPG available with jpg_quality (95 default).
• skip_if_exists - True and it won't overwrite existing files, though it still reports their paths.
• alpha_mask / invert_alpha / save_alpha_separately - the alpha pipeline. alpha_mask bakes a mask into the saved file's alpha (white = opaque), invert_alpha flips polarity, and save_alpha_separately also writes a greyscale _alpha.png alongside.

Outputs: saved_count (INT), filepaths (STRING), and status (STRING) - so you can hand the paths off to the next node if something downstream needs to know where the files landed.

The workflow shape

The pack's recommended PBR pipeline is the perfect illustration: BD Derive PBR Maps outputs its seven maps, all of them feed input_1..input_7, the labels field reads albedo\nnormal\nroughness\nmetallic\nao\norm\narm, and one run writes all seven files into the context-resolved folder. One node doing the job of seven. If you're not on the PBR pipeline, it's just as good for "save this character's head, hands, clothes and shoes in one shot."

Installing it

Part of BizaNator/ComfyUI-BrainDead (BrainDeadGuild). ComfyUI Manager → search "BrainDead" → install, or:

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

Restart; it's under 🧠BrainDead/Cache.

Gotchas

The label alignment is the thing that bites. Labels align with wired inputs after empty slots are skipped - so if you wire input_1 and input_3 but leave input_2 empty, the labels line up with input_1 and input_3, not with the slot numbers. It's sensible once you know it, infuriating until you do. Second: this node saves in one run - it's the right tool when the downstream is just saving. If the downstream is more complex (per-part upscaling before save), the pack's BD For Each Run iterator is the one that emits per-item labels across runs. And remember it's an output node: it runs on every queue, so keep skip_if_exists in mind if you're re-running the same workflow repeatedly.