EasyUse/PreSamplinggenerated
PreSampling (Cascade)(cascadeSettings)
PreSampling (Cascade) is a pipeline-configuration node for sampling with Stable Cascade models.
cascade Settings
pipe
image_to_latent_c
latent_c
pipe
encode_vae_name
COMBO
decode_vae_name
COMBO
steps
20
cfg
4
sampler_name
euler_ancestral
scheduler
simple
denoise
1.00
seed
INT
Easy Use
Description
PreSampling (Cascade) is a pipeline-configuration node for sampling with Stable Cascade models. It accepts a PIPE_LINE, updates it with the requested sampling parameters, selects the VAE components used for the Cascade encode/decode stages, and optionally attaches a latent-C conditioning image or existing latent. It does not perform sampling itself; downstream sampler nodes use the returned PIPE_LINE.
Inputs
| Input | Type | Default | Description |
|---|---|---|---|
pipe | PIPE_LINE | (required) | The pipeline containing the model, conditioning, existing latents, and sampler context. This node reads the pipeline and returns it with updated settings. |
encode_vae_name | COMBO | (required) | Selects the VAE used to encode image_to_latent_c into latent C. Available VAE names depend on the loaded environment. |
decode_vae_name | COMBO | (required) | Selects the VAE used to decode latent C back into an image. In Stable Cascade this can be a different VAE from the encode VAE. |
steps | INT | 20 | Number of denoising steps. Allowed range: 1 to 10000. |
cfg | FLOAT | 4.0 | Classifier-free guidance scale. Allowed range: 0 to 100. |
sampler_name | COMBO | euler_ancestral | The sampler to use for denoising. |
scheduler | COMBO | simple | The scheduler to use with the selected sampler. |
denoise | FLOAT | 1.0 | How much of the existing latent to denoise. Allowed range: 0 to 1, step 0.01. Use 1.0 to sample fully from noise; lower values preserve more of the original latent content. |
seed | INT | (no fixed default) | Random seed for reproducible sampling. |
image_to_latent_c | IMAGE | (optional) | An image to encode into latent C using encode_vae_name. Useful when you want to provide image-derived latent-C conditioning. |
latent_c | LATENT | (optional) | A ready-made latent C to use directly. Useful when you already have the encoded latent and want to skip the image-encoding step. |
Outputs
| Output | Type | Description |
|---|---|---|
pipe | PIPE_LINE | The same pipeline with the pre-sampling settings applied, including the VAE selections and any optional latent-C input, ready for downstream sampler nodes. |
Usage Notes
- Stable Cascade workflows can use separate VAEs for encoding images into latents and decoding latents back into images, so
encode_vae_nameanddecode_vae_nameshould be set to match the cascade model components you are using. - Use
image_to_latent_cwhen you need to create latent C from an input image. Uselatent_cwhen you already have the latent-C representation and do not need to encode an image. - This node only configures the pipeline; connect its
PIPE_LINEoutput to a downstream sampler that supports pipeline-based sampling.
Comments
Sign in with GitHub to join the discussion.