2.5 KiB
2.5 KiB
TT KSampler (Advanced) TT_KSamplerAdvancedNode
This description was generated with AI assistance. If you spot any errors, please create an issue on GitHub.
Extended KSampler with fine-grained control over noise, denoising range, and step boundaries. Useful for multi-stage workflows and img2img.
Inputs
| Parameter | Type | Required | Description |
|---|---|---|---|
model |
MODEL | Yes | Diffusion model |
positive |
CONDITIONING | Yes | Positive conditioning |
negative |
CONDITIONING | Yes | Negative conditioning |
latent |
LATENT | Yes | Input latent tensor |
add_noise |
BOOLEAN | Yes | Random Noise — use random noise; Zero Noise — use zeros |
full_denoise |
BOOLEAN | Yes | Complete — force full denoise; Partial — stop early |
denoise |
FLOAT | Yes | Denoising strength, default 1.0 |
seed |
INT | Yes | Generation seed |
steps |
INT | Yes | Total number of sampling steps, default 30 |
cfg |
FLOAT | Yes | Guidance scale, default 3.0 |
start_step |
INT | No | Step to start sampling from, default 0 |
last_step |
INT | No | Step to stop sampling at, default 10000 |
sampler_name |
COMBO | Yes | Sampler algorithm |
scheduler |
COMBO | Yes | Noise schedule |
preview_latent |
BOOLEAN | No | Show step-by-step latent preview during sampling |
Outputs
| Parameter | Type | Description |
|---|---|---|
LATENT |
LATENT | Sampled latent tensor |
Usage
Use start_step and last_step to run only a portion of the schedule — useful for chaining two samplers in a draft/refine workflow. Set add_noise
to Zero Noise when passing a pre-noised latent from a previous stage.