TT SDXL Workflow Settings

Somewhere in every SDXL workflow, the same five values live in a row: seed, steps, CFG, sampler, scheduler. Native ComfyUI scatters them across every KSampler you add, so the moment you have two samplers (hires pass, anyone?) you're editing the same settings in two places. TT SDXL Workflow Settings is the card that holds them once, hands them out bundled in a WORKFLOW_CONFIG, and lets the pack's context-aware nodes pick them up without you ever re-entering them.

It's a pure settings node - no loading, no sampling, no side effects. It packages seed, steps, cfg, sampler_name, and scheduler into a config object and also exposes each one as its own typed output (SEED, STEPS, CFG, SAMPLER_NAME, SCHEDULER). The config object is what feeds the context pipeline: TT KSampler (Context) reads the workflow_config out of a context and runs the sampler with these values. The individual outputs are your escape hatch - plain INT/FLOAT/STRING types that plug into native nodes, so you can use this as a tidy "settings panel" even if you never touch the context system.

The defaults are SDXL-standard and worth internalizing: steps 30, cfg 7.0. SDXL runs happily at CFG 6–8 (unlike Flux, where CFG barely exists and guidance does the work - this node has no guidance field for exactly that reason), and 30 steps with euler or dpmpp is the boring, reliable starting point. sampler_name and scheduler give you the full 44-sampler / 9-scheduler lists from ComfyUI, not a trimmed subset.

Inputs that matter:

• seed - default 0, which means "the same image every time until you change it." Randomize it for variety or wire a fixed value for reproducibility.
• cfg - the SDXL quality dial; 7 is stock, 8–9 for a stronger prompt lock, lower for more freedom.
• steps - 30 is comfortable; you can push toward 40 for finicky styles or drop to 25 with a good scheduler.

Install is the pack standard:

cd ComfyUI/custom_nodes
git clone https://github.com/tenser-tensor/ComfyUI-TenserTensor

or search "TenserTensor" in ComfyUI Manager and restart. Only declared deps are gguf and kornia.

The one trap is the walled-garden rule that runs through the whole pack: TT_WORKFLOW_CONFIG is proprietary to TenserTensor. Wire it into a native node and you get a type mismatch. That's not a bug - it's the pack keeping its context types to itself. If you don't want the context system, just use the individual outputs, which are all standard types and work anywhere.

Also worth knowing: this is the V1 node and the pack labels it deprecated as it migrates to ComfyUI's API V3 (TT_SdxlWorkflowSettingsNode is the maintained replacement). It works fine today, and for a graph that's already running TenserTensor context nodes, it's still the natural settings card.