📁 Bawk Batch Processor

If you've ever sat there clicking "Queue" fifty times because you're A/B testing five prompts across ten seeds, this is the node that rescues you. Bawk Batch Processor is the automation end of the Bawk Nodes pack: it reads one row from a CSV or JSON file and hands that row's settings to the rest of your workflow. You bump batch_index, hit Queue, and the same node chain produces the next variation. It won't auto-queue itself - this isn't a scheduler - but it turns "one workflow per prompt" into "one workflow, many rows."

How it works

The node reads your file with Python's csv/json modules and auto-detects the format from the extension, or you can force it with file_format. batch_index (0-based) picks the row. Each row can carry prompt, seed, steps, guidance, and resolution; anything missing falls back to default_resolution, default_steps, and default_guidance. The batch_info output tells you exactly where you are ("Processing item 3/20 from prompts.csv"), which you'll appreciate when you come back after a coffee and can't remember.

preview_only is the trick that saves you a wasted run: flip it on and the node shows the file contents instead of generating. Check your file before you commit to a 20-minute batch.

The inputs that matter

The CSV columns are what you actually set up, per the README:

prompt,seed,steps,guidance,resolution
"beautiful sunset",12345,30,3.5,"Instagram Square - 1024x1024 - 1.0MP"
"city at night",67890,25,4.0,"FHD 16:9 - 1920x1080 - 2.1MP"

• batch_file - the absolute path on the machine running ComfyUI.
• batch_index - which row to process. Start at 0.
• preview_only - verify before you commit.

Wiring it in

The pack's batch chain is Bawk Batch Processor → Bawk Wildcard Encoder → Bawk Sampler → Bawk Image Saver. prompt feeds the encoder's prompt field, and seed/steps/guidance/resolution feed the sampler's matching sockets. Since the encoder understands {wildcard|syntax} prompt syntax, you can get combinatorial variation for free on top of your CSV rows.

Install & gotchas

Install comes with the pack - ComfyUI Manager → search "Bawk Nodes" → Install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/juddisjudd/ComfyUI-BawkNodes.git

Restart ComfyUI afterward. The pack's requirements.txt is just torch, torchvision, safetensors, numpy and requests - everything stock ComfyUI already pulls in, so there's no extra dependency dance for this node.

Where people get burned:

• The path is server-side. If ComfyUI runs on another machine, a C:\... path from your laptop won't resolve. The file must exist on the box running ComfyUI.
• Column names must match exactly. Resolution instead of resolution silently falls back to the default rather than erroring.
• batch_index past the last row returns an error string in every output - you'll see it instantly, so don't panic, just reset the index.

That's the whole trick: it's a dumb, reliable settings-forwarder, and that's exactly what batch automation needs.