basic_Ksampler_mid

basic_Ksampler_full assumes you're running the whole Apt_Preset pipeline and always hands it a RUN_CONTEXT. basic_Ksampler_mid is the version for when you're not - context here is optional, and every model-related input (model, positive, negative, latent, vae, clip) is optional too, wired individually if you want, or left for the context to supply if you're using one. It's the middle ground the name promises: less rigid than the full context-driven node, more compact than wiring a stock KSampler plus a separate preview node plus a separate VAE decode by hand.

What it is

Only three fields are actually required: seed, denoise (0-1, default 1), and image_output (the Hide/Preview/Save/Hide/Save enum this whole node family shares, letting you skip a separate Preview/Save Image node). Everything else - context, model, positive, negative, latent, vae, clip, image - is optional, which means you can drop this into a graph that's using this pack's context pipeline or a graph that's wired the traditional stock-ComfyUI way, model/clip/conditioning connected node-to-node.

The inputs and outputs that matter

Set seed and denoise - the two knobs you'll touch on essentially every run - and image_output if you want an inline preview or save. If you're feeding it a context, that supplies whatever you don't wire directly; if you're not, wire model, positive, negative, and latent the conventional way, same as a stock KSampler. Notice what's absent from the required list compared to basic_Ksampler_full: no steps, cfg, sampler, or scheduler fields exposed here at all - this node leans on the context (or its internal defaults) for those rather than exposing them as its own widgets, which is the real trade you're making for the compactness. Outputs mirror basic_Ksampler_full: context, image, model, positive, negative, latent, vae, clip.

Which of the five samplers to actually reach for

This pack ships five KSampler variants and it's genuinely not obvious at a glance which one fits your graph. As a rule of thumb: basic_Ksampler_full if you're all-in on the context pipeline and want every knob exposed; basic_Ksampler_mid (this one) if you're mixing context-driven and manually-wired inputs, or just want a lighter node than the full version; basic_Ksampler_adv specifically for staged/refiner sampling with start_at_step/end_at_step; basic_Ksampler_custom only once you're assembling atomic NOISE/GUIDER/SAMPLER/SIGMAS objects yourself; and basic_Ksampler_low_gpu when the problem you're solving is specifically a VAE-decode memory spike, not the sampling steps themselves. If you're not sure which applies, this middle node is the safest default to start with.

Installing it

Search "ComfyUI-Apt_Preset" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git

Run install.bat on Windows (pip install -r requirements.txt on Linux/Mac), restart. Nothing extra beyond your usual model/VAE files.

Common issues

Because so much here is optional, the failure mode is usually a silent one: forget to wire either a context carrying the sampling settings or enough of the individual overrides, and the node either falls back to internal defaults you didn't intend or errors on a genuinely missing piece - check exactly which of model/positive/negative/latent you actually connected before assuming the node itself is broken. If you need fine control over steps, CFG, sampler, or scheduler and this node isn't exposing them the way you expect, that's a sign you actually want basic_Ksampler_full instead - this one is deliberately the lighter option, not a strict superset.