Files

36 lines
1.8 KiB
Markdown

## TT KSampler *TT_KSamplerNode*
*This description was generated with AI assistance. If you spot any errors, please create an issue on GitHub.*
Standard KSampler node with built-in latent preview support. Accepts model, conditioning, and latent as separate inputs.
### 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 |
| `seed` | INT | Yes | Generation seed |
| `steps` | INT | Yes | Number of sampling steps, default 30 |
| `cfg` | FLOAT | Yes | Guidance scale, default 3.0 |
| `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
Connect `MODEL`, `CONDITIONING`, and `LATENT` outputs from loader and encoder nodes. Enable `preview_latent` to see intermediate results during
sampling.
---
[View on GitHub](https://github.com/tenser-tensor/ComfyUI-TenserTensor)