Operating Mode v2

Operating Mode v2 is where you tell the Searge workflow what kind of job it's doing - plain text-to-image, image-to-image, or inpainting - plus how it should treat your prompts and how many images to make. It's one of the few Searge nodes that isn't deprecated, so if you're actually running the current v4.x workflow, this is a node you'll touch.

Some background: SeargeSDXL is the SDXL base-plus-refiner workflow pack that Searge put out when SDXL 1.0 launched in mid-2023. It was one of the first "everything in one workflow" packs, and people leaned on it hard in those early days. The v4 rewrite replaced the old scatter of loose parameter wires with a single data stream - a SRG_DATA_STREAM bus that threads through the config nodes. This node sits on that bus near the front and stamps the mode onto it.

How it works

You set the mode here, and everything downstream reads it off the data stream instead of you rewiring anything. Switch to image-to-image and the graph starts sampling from your source image; switch to in-painting and it also honors the mask. The prompting_mode dropdown decides which of your prompt boxes actually get used - handy when you want to quickly ignore the secondary or style prompts without deleting the text. Because it all rides one wire, flipping modes is a dropdown change, not a re-plumbing job.

The data input is optional because this node can either start a fresh stream or extend one that an earlier node began. Either way it emits an updated data stream out the other side.

The inputs and outputs that matter

• workflow_mode (enum: none, text-to-image, image-to-image, in-painting) - the big one. This is the mode switch the whole workflow pivots on.
• prompting_mode (enum, 7 options like default - all prompts, main and neg. only, ignore empty, ignore all) - which prompt fields are honored. Great for A/B-ing whether your style prompt is helping.
• batch_size (INT, default 1, max 4) - how many images per run.
• data in and out (SRG_DATA_STREAM) - the bus; wire it through in order.

How to install it

ComfyUI Manager: search SeargeSDXL, install, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/SeargeDP/SeargeSDXL.git, then restart. Manual installers need python -m pip install opencv-python in ComfyUI's Python environment first, per the README, or the nodes won't load.

Because this drives a full generation workflow, you'll want the pack's models installed too - SDXL base and refiner checkpoints at minimum, the fp16-fix VAE, and for inpainting/img2img the usual supporting files. The README lists every download with direct links.

Common issues

The classic Searge failure is a version mismatch - an old workflow JSON against new nodes, or the reverse. The README is blunt about it: always load the latest workflow file with the latest node version. Do that before you debug anything else.

For inpainting specifically, setting the mode here is only half the job; you still have to feed a source image and a mask in (that's what the Image Adapter node is for), and denoise matters a lot - too high and the whole region regenerates from scratch, too low and nothing changes. Start moderate and adjust.

And the honest big-picture note: the Searge workflow is built around SDXL's refiner, which the community has mostly abandoned for single fine-tuned checkpoints. This node still works fine, but if you're new, know that you're learning a somewhat historical setup rather than the current default.