This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHubThe LTXVScheduler node generates sigma values for custom sampling processes. It calculates noise schedule parameters based on the number of tokens in the input latent and applies a sigmoid transformation to create the sampling schedule. The node can optionally stretch the resulting sigmas to match a specified terminal value.
Inputs
| Parameter | Data Type | Required | Range | Description |
|---|---|---|---|---|
steps | INT | Yes | 1-10000 | Number of sampling steps (default: 20) |
max_shift | FLOAT | Yes | 0.0-100.0 | Maximum shift value for sigma calculation (default: 2.05) |
base_shift | FLOAT | Yes | 0.0-100.0 | Base shift value for sigma calculation (default: 0.95) |
stretch | BOOLEAN | Yes | True/False | Stretch the sigmas to be in the range [terminal, 1] (default: True) |
terminal | FLOAT | Yes | 0.0-0.99 | The terminal value of the sigmas after stretching (default: 0.1) |
latent | LATENT | No | - | Optional latent input used to calculate token count for sigma adjustment |
latent parameter is optional. When not provided, the node uses a default token count of 4096 for calculations.
Outputs
| Output Name | Data Type | Description |
|---|---|---|
sigmas | SIGMAS | Generated sigma values for the sampling process |