ComfyUI Node
Dazzle KSampler Advanced (DazzleNodes)
A ComfyUI node in RES4LYF/samplers with 30 inputs and 1 output.
Dazzle KSampler Advanced (DazzleNodes)
- guides
- automation
- options
- sampler
◄noise_type_sdegaussian►
◄noise_type_sde_substepgaussian►
◄noise_mode_sdehard►
◄noise_mode_sde_substephard►
◄overshoot_modehard►
◄overshoot_mode_substephard►
◄eta0.50►
◄eta_substep0.50►
◄overshoot0.00►
◄overshoot_substep0.00►
◄noise_scaling_weight0.00►
◄noise_boost_step0.00►
◄noise_boost_substep0.00►
◄noise_anchor1.00►
◄s_noise1.00►
◄s_noise_substep1.00►
◄d_noise1.00►
◄momentum1.00►
◄noise_seed_sde-1►
◄latent_roleauto►
◄sampler_nameres_2m►
◄implicit_typepredictor-corrector►
◄implicit_type_substepspredictor-corrector►
◄implicit_steps0►
◄implicit_substeps0►
◄bongmathtrue►
◄extra_options►
CategoryRES4LYF/samplers
Inputs (30)
| Name | Type | Default | Description |
|---|---|---|---|
| noise_type_sde | COMBO | gaussian | 25 options: none, brownian, gaussian, gaussian_backwards, laplacian, perlin, +19 |
| noise_type_sde_substep | COMBO | gaussian | 25 options: none, brownian, gaussian, gaussian_backwards, laplacian, perlin, +19 |
| noise_mode_sde | COMBO | hard | How noise scales with the sigma schedule. Hard is the most aggressive, the others start strong and drop rapidly. |
| noise_mode_sde_substep | COMBO | hard | How noise scales with the sigma schedule. Hard is the most aggressive, the others start strong and drop rapidly. |
| overshoot_mode | COMBO | hard | How step size overshoot scales with the sigma schedule. Hard is the most aggressive, the others start strong and drop rapidly. |
| overshoot_mode_substep | COMBO | hard | How substep size overshoot scales with the sigma schedule. Hard is the most aggressive, the others start strong and drop rapidly. |
| eta | FLOAT | 0.50-100–100 | Calculated noise amount to be added, then removed, after each step. |
| eta_substep | FLOAT | 0.50-100–100 | Calculated noise amount to be added, then removed, after each step. |
| overshoot | FLOAT | 0.00-100–100 | Boost the size of each denoising step, then rescale to match the original. Has a softening effect. |
| overshoot_substep | FLOAT | 0.00-100–100 | Boost the size of each denoising substep, then rescale to match the original. Has a softening effect. |
| noise_scaling_weight | FLOAT | 0.00-100–100 | Set to positive values to create a sharper, grittier, more detailed image. Set to negative values to soften and deepen the colors. |
| noise_boost_step | FLOAT | 0.00-100–100 | Set to positive values to create a sharper, grittier, more detailed image. Set to negative values to soften and deepen the colors. |
| noise_boost_substep | FLOAT | 0.00-100–100 | Set to positive values to create a sharper, grittier, more detailed image. Set to negative values to soften and deepen the colors. |
| noise_anchor | FLOAT | 1.00-100–100 | Typically set to between 1.0 and 0.0. Lower values cerate a grittier, more detailed image. |
| s_noise | FLOAT | 1.00-10000–10000 | Adds extra SDE noise. Values around 1.03-1.07 can lead to a moderate boost in detail and paint textures. |
| s_noise_substep | FLOAT | 1.00-10000–10000 | Adds extra SDE noise. Values around 1.03-1.07 can lead to a moderate boost in detail and paint textures. |
| d_noise | FLOAT | 1.00-10000–10000 | Downscales the sigma schedule. Values around 0.98-0.95 can lead to a large boost in detail and paint textures. |
| momentum | FLOAT | 1.00-10000–10000 | Accelerate convergence with positive values when sampling, negative values when unsampling. |
| noise_seed_sde | INT | -1-1–18446744073709550000 | — |
| latent_role | COMBO | auto | ADVISORY: this node returns a SAMPLER object and does not consume a latent input directly; the dispatch decision is made by the downstream node that consumes the SAMPLER. The widget is shown for UI consistency with the other Dazzle sampler nodes. |
| sampler_name | COMBO | res_2m | 119 options: none, res_2m, res_3m, dpmpp_2m, dpmpp_3m, abnorsett_2m, +113 |
| implicit_type | COMBO | predictor-corrector | 4 options: rebound, retro-eta, bongmath, predictor-corrector |
| implicit_type_substeps | COMBO | predictor-corrector | 4 options: rebound, retro-eta, bongmath, predictor-corrector |
| implicit_steps | INT | 00–10000 | — |
| implicit_substeps | INT | 00–10000 | — |
| bongmath | BOOLEAN | true | — |
| guidesopt | GUIDES | — | |
| automationopt | AUTOMATION | — | |
| extra_optionsopt | STRING | — | |
| optionsopt | OPTIONS | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| sampler | SAMPLER | — |