HybridSamplers for ComfyUI
Custom samplers and schedulers for ComfyUI
HybridSamplers for ComfyUI
HybridSamplers is a ComfyUI custom node extension that introduces new experimental samplers and schedulers designed to give more control and flexibility over the diffusion process.
With HybridSamplers, you can explore alternative numerical methods, dynamic scheduling strategies, and experimental noise shaping — useful for research, experimentation, or pushing creative generation beyond standard defaults.
✨ Features
Custom Samplers
-
AdaptiveEuler – Euler-based sampler with adaptive scaling.
-
DynamicLangevin – Langevin dynamics with noise adaptation.
-
StochasticRungeKutta – Runge–Kutta method with stochastic jitter.
-
TemporalSampling – Time-aware blending between diffusion steps.
-
SpatialSampling – Spatially perturbed sampling for added detail.
-
Quantized – Rounds latents to discrete bins for a quantized look.
-
Anisotropic – Applies anisotropic noise for directional detail.
-
MultiDimensional – Hybrid Euler + Heun blending.
Images generated at 15 steps with cfg 7.0 and seed 42
Custom Schedulers
- AdaptiveTime – Time-decaying schedule.
- DynamicSchedule – Cosine/linear interpolation scaling.
- VariableStep – Adjustable step ranges.
- ProgressiveDecay – Interval-based decay factor.
- AdaptiveExponential – Growth + saturation scaling.
- FractalTime – Fractal-based scaling factor.
- TemporalGradient – Gradient-magnitude based schedule.
- MemoryAware – Retains partial state across steps.
- MultiObjective – Weighted averaging.
- ResourceConstrained – Efficiency-aware scaling.
- DynamicWindow – Adaptive vs fixed step scaling.
- MultiAgent – Multi-agent coordination scaling.
🔧 Installation
Clone into your ComfyUI/custom_nodes folder:
cd ComfyUI/custom_nodes
git clone https://github.com/azazeal04/ComfyUI-HybridSamplers.git
Install from manager
Search for HybridSamplers for ComfyUI in the manager and install it.
Restart ComfyUI and the new samplers and schedulers will be available in dropdown menus.
🛠 Development
- Python ≥ 3.10
- Torch ≥ 2.0
- Requires ComfyUI installed and working.
⚠️ Notes
- These samplers/schedulers are experimental and may produce noisy or blurry results.
- Designed for testing new ideas in diffusion sampling, not guaranteed for production use.
- You can enable debug logs in the console to see when a custom sampler/scheduler is active.
📜 License
This project is licensed under the MIT License – see LICENSE for details.