MD: Hybrid Scheduler (Basic)

The Advanced Hybrid Scheduler in this pack is a control room with 30 knobs. This is the same engine with five. If you want the Karras/Bong/AYS scheduling math without tripping over beta_beta and blend_factor, Basic is the one you'll actually reach for.

It does the important thing the stock ComfyUI scheduler doesn't: it reads your model input to auto-detect the model's real sigma min/max, then generates a schedule tuned to that range. Then it applies your chosen curve.

The inputs

• model - your checkpoint. The node pulls model_sampling off it for the sigma bounds.
• steps - 30 by default; the tooltip suggests 20–30.
• mode - the curve, same 14-mode list as the Advanced node: karras_rho (the default, right for SD 1.5/SDXL), bong_tangent (the audio-oriented one), ays, poly, beta, simple, linear_quadratic, sgm_uniform, and friends.
• denoise - 1.0 for full generation, lower for img2img or refinement. The node keeps the end of the schedule, which is the correct way to do partial denoise.
• rho - curve steepness for the Karras/poly family. 7.0 is standard; lower flattens it.

Outputs

SIGMAS (wire into a sampler that accepts a sigma schedule - SamplerCustom, or this pack's samplers like MD_AMED_Sampler) and IMAGE - a matplotlib plot of the curve, handy for seeing what the schedule actually looks like before you burn GPU time on it.

Installing it

Same pack as everything: 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

The plot output needs matplotlib, which the pack's requirements installs.

Common issues

Don't go crazy with rho on flow-matching checkpoints (Flux and friends). Aggressive Karras reshaping is a correction on DDPM-style SD 1.5/SDXL models but a distortion on near-straight flow trajectories - if your output degrades, switch mode to simple, sgm_uniform, or linear_quadratic rather than fighting the curve. And if the plot comes back blank, matplotlib is missing from your environment - pip install matplotlib and restart.