MD: Hybrid Scheduler (Lite)

If the Advanced Hybrid Scheduler is the control room and Basic is the five-knob console, Lite is the three-button kiosk. model, steps, denoise. That's it. It always runs the pack's karras_rho curve at rho 7 - the standard, boring, reliable choice - and hands you a clean SIGMAS schedule.

That's the whole appeal, and it's a real one: the most useful thing this node family does is read the model's actual sigma range instead of assuming, and Lite does that with zero configuration surface. You drop your checkpoint in, set steps (20 by default), and get a Karras-style schedule tuned to that model.

When to use it

Honestly, Lite is the one I'd grab first as a drop-in scheduler experiment. It's also the version to steal for a shared workflow, because there's nothing to misconfigure. Wire SIGMAS into a sampler that accepts a sigma input - SamplerCustom, or this pack's own samplers - and compare against ComfyUI's built-in scheduler on a fixed seed. The difference is usually subtle on images and more audible on audio, which is where this pack's schedule math genuinely shines.

One caveat, straight from the KB: karras_rho is the right default for DDPM-style checkpoints (SD 1.5, SDXL and their finetunes), but flow-matching models like Flux dislike aggressive schedule reshaping. On those, you want Basic so you can switch to simple, sgm_uniform, or linear_quadratic. If you're on a flow-matching checkpoint, Lite is the wrong flavor.

Installing it

Part of MD Nodes (MDMAchine/ComfyUI_MD_Nodes). ComfyUI Manager → search MD_Nodes → Install, restart, or:

cd path/to/ComfyUI/custom_nodes
git clone https://github.com/MDMAchine/ComfyUI_MD_Nodes.git
cd ComfyUI_MD_Nodes
pip install -r requirements.txt

Restart ComfyUI. Standard pack context: Python 3.10+, GPL v3, pure Python. Unlike the Basic and Advanced variants, Lite outputs only SIGMAS - no matplotlib plot - so it has no extra dependency beyond the pack itself.