Pipe Out Sampler Settings (Simple)
Simple sampler settings, plus the upscaler, on one pipe
- pipe
- pipe
- sampler
- scheduler
- steps
- cfg
- flux_guidance
- upscale_model
- scale_by
The name undersells it. "Pipe Out Sampler Settings (Simple)" doesn't just unpack the basics - it's one of two variants in this family that also carries an upscale_model and a scale_by. So the bundle it's splitting isn't just "how do I sample," it's "how do I sample, then upscale." That makes it the natural fit for a hires-fix-style workflow where the KSampler and the upscale model live on the same logical branch.
Output list: sampler, scheduler, steps, cfg, flux_guidance, upscale_model, scale_by, plus the pipe passed through. The first five are the sampler controls you know; the last two are what earn this node its keep. upscale_model is the model object itself - the node carries a loaded upscaler through the pipe, which is handy because upscale models aren't plain numbers and are annoying to route by hand. scale_by is the multiplier you feed to the upscale operation.
The one real quirk to understand: this pipe has a flux_guidance field even though "Simple" suggests a generic setup. That's the pack keeping a common pipe shape - the value rides along whether your checkpoint is Flux or SDXL. On SDXL it's simply not consumed; on Flux it's the guidance knob. If the pipe's counterpart is a "Simple" Pipe In, feed flux_guidance when you're on a flow model and ignore it otherwise.
Mechanically it's the usual RvTools deal: one pipe input, a passthrough pipe output, and each bundled value split to its own socket via a straightforward tuple unpack. No file I/O, no state.
When would you actually reach for this one? When your workflow does sample-then-upscale and you want both the sampler settings and the upscale plan to travel as one coherent bundle - so you can't accidentally feed the KSampler one step count and the upscaler another. It's also the pair you'd pick if you're moving between checkpoint families, since sampler/scheduler/steps/guidance all arrive in one place ready to be rewired.
Installing it
Ships in ComfyUI-RvTools_v2. ComfyUI Manager: search "ComfyUI-RvTools_v2". Or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2
Restart ComfyUI. Requirements (torch, numpy, Pillow, opencv-python, pilgram, pynvml, piexif) install via Manager. No model downloads, no keys.
Where people get burned
- Pipe shape, again. This unpacker expects the seven-field Simple pipe (with upscale fields). Feed it a Small pipe and you'll get a tuple
ValueErroronly after you hit queue. upscale_modelneeds a real loaded model. If the field arrives asNone, the upscale leg of the workflow fails downstream - check the Pipe In node feeding it, not this node.- Deprecation. The RvTools v2 pack is unmaintained and the README points to ComfyUI_Eclipse. Runs fine as installed; just not where new work should live.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| pipe | pipe | — |
Outputs (8)
| Name | Type | Description |
|---|---|---|
| pipe | pipe | — |
| sampler | * | — |
| scheduler | * | — |
| steps | INT | — |
| cfg | FLOAT | — |
| flux_guidance | FLOAT | — |
| upscale_model | UPSCALE_MODEL | — |
| scale_by | FLOAT | — |