Vito Sansevero
ab23992c29
- Add code quality tools: flake8, mypy, black, pre-commit - Add .gitattributes for line ending consistency - Add .secrets.baseline for secret scanning - Update GitHub workflows for better CI/CD - Update documentation formatting and examples - Add CLAUDE.md for AI assistant guidance - Add scripts directory for automation tools - Update project configuration in pyproject.toml - Improve type hints and code formatting across all modules - Update test configurations and fixtures
5.7 KiB
5.7 KiB
Sampler Combo Documentation
Overview
The Sampler Combo is a unified ComfyUI node that combines sampler, scheduler, steps, and CFG settings into a single interface. It reduces workflow complexity while ensuring compatible parameter combinations and providing optimization recommendations.
Features
🎯 Unified Configuration
- Single node for all sampling parameters
- Compatible sampler + scheduler combinations
- Optimized steps and CFG recommendations
- Reduced workflow complexity
🧠 Smart Recommendations
- Scheduler suggestions based on selected sampler
- Optimal steps range for each sampler
- CFG scale recommendations
- Compatibility warnings and suggestions
✅ Built-in Validation
- Parameter validation and sanitization
- Error handling with safe defaults
- Performance optimization hints
- Real-time compatibility checking
📊 Analysis Tools
- Combo configuration analysis
- Performance assessment
- Optimization recommendations
- Compatibility scoring
Node Interface
Inputs
- sampler_name: Dropdown with available sampling algorithms
- scheduler: Dropdown with compatible schedulers
- steps: Integer slider (1-100 steps)
- cfg: Float slider (0.0-20.0 CFG scale)
Outputs
- sampler_name: Selected sampler algorithm
- scheduler: Selected scheduler algorithm
- steps: Number of sampling steps
- cfg: CFG scale value
Available Samplers
Primary Samplers
| Sampler | Type | Speed | Quality | Best For |
|---|---|---|---|---|
| euler | Deterministic | Fast | Good | General use |
| euler_ancestral | Stochastic | Fast | Good | Creative variation |
| heun | Higher-order | Medium | Better | Quality focus |
| dpm_2 | Multi-step | Medium | Good | Balanced |
| dpm_2_ancestral | Stochastic | Medium | Good | Creative quality |
| lms | Linear | Fast | Good | Simple scenes |
| dpm_fast | Optimized | Very Fast | Good | Speed priority |
| dpm_adaptive | Adaptive | Variable | Best | Automatic tuning |
Advanced Samplers
| Sampler | Type | Speed | Quality | Best For |
|---|---|---|---|---|
| dpmpp_2s_ancestral | Advanced | Medium | Better | High quality |
| dpmpp_2m | Optimized | Fast | Better | Speed + quality |
| dpmpp_2m_sde | Stochastic | Medium | Best | Maximum quality |
| dpmpp_3m_sde | Latest | Medium | Best | Cutting edge |
| ddim | Classic | Fast | Good | Compatibility |
| uni_pc | Unified | Fast | Better | Efficiency |
Available Schedulers
Linear Schedulers
- normal: Standard linear schedule
- linear: Basic linear distribution
- sgm_uniform: Uniform distribution
Advanced Schedulers
- karras: Karras noise schedule (recommended)
- exponential: Exponential decay
- polyexponential: Polynomial exponential
- beta: Beta distribution schedule
Specialized Schedulers
- cosine: Cosine annealing schedule
- simple: Simplified schedule
- ddim_uniform: DDIM uniform schedule
- laplace: Laplace distribution
Optimization Guidelines
Recommended Combinations
Speed Optimized
Sampler: euler or dpm_fast
Scheduler: normal or simple
Steps: 15-25
CFG: 6.0-8.0
Quality Optimized
Sampler: dpmpp_2m_sde or dpmpp_3m_sde
Scheduler: karras
Steps: 25-35
CFG: 7.0-9.0
Balanced
Sampler: dpmpp_2m or heun
Scheduler: karras or normal
Steps: 20-30
CFG: 7.0-8.5
Steps Recommendations
| Sampler Type | Min Steps | Optimal | Max Steps |
|---|---|---|---|
| Fast (euler, dpm_fast) | 10 | 20 | 30 |
| Standard (heun, dpm_2) | 15 | 25 | 40 |
| Advanced (dpmpp_*) | 20 | 30 | 50 |
| Adaptive | 10 | 25 | 100 |
CFG Scale Guidelines
| Content Type | CFG Range | Recommended |
|---|---|---|
| Photorealistic | 5.0-8.0 | 7.0 |
| Artistic/Stylized | 7.0-12.0 | 9.0 |
| Abstract/Creative | 8.0-15.0 | 11.0 |
| Text/Details | 10.0-20.0 | 13.0 |
Usage Examples
Basic Configuration
sampler_name: euler
scheduler: normal
steps: 20
cfg: 7.0
High Quality Setup
sampler_name: dpmpp_2m_sde
scheduler: karras
steps: 30
cfg: 8.0
Speed Priority
sampler_name: dpm_fast
scheduler: simple
steps: 15
cfg: 6.5
Advanced Features
Compatibility Analysis
The node provides real-time analysis of parameter compatibility:
- Scheduler compatibility with selected sampler
- Steps optimization for sampler type
- CFG scale recommendations
- Performance impact assessment
Error Handling
- Invalid samplers default to 'euler'
- Invalid schedulers default to 'normal'
- Out-of-range steps clamped to valid range
- Invalid CFG values sanitized to safe defaults
Performance Tips
- Use Karras scheduler for most samplers (better quality)
- Start with 20-30 steps for most use cases
- Keep CFG 6.0-9.0 for realistic images
- Try dpmpp_2m for best speed/quality balance
- Use euler for fastest generation
Troubleshooting
Common Issues
- Slow generation: Try euler or dpm_fast samplers
- Poor quality: Increase steps or try dpmpp_2m_sde
- Overcooked images: Lower CFG scale
- Underdetailed: Increase CFG or steps
- Artifacts: Try karras scheduler or different sampler
Performance Optimization
- GPU Memory: Lower steps if running out of VRAM
- Speed: Use euler + normal + 15-20 steps
- Quality: Use dpmpp_2m_sde + karras + 25-30 steps
- Compatibility: Stick to euler/heun for broad model support
Integration
The Sampler Combo node outputs are compatible with all standard ComfyUI sampling nodes:
- KSampler
- KSamplerAdvanced
- Custom sampling workflows
- Upscaling pipelines
- Img2img workflows
Connect the outputs directly to your sampling node inputs for streamlined configuration.