Pipe Edit(pipeEdit)
Pipe Edit is an all-in-one pipeline editing node for EasyUse pipes.
pipe Edit
Description
Pipe Edit is an all-in-one pipeline editing node for EasyUse pipes. It accepts a PIPE_LINE and optional direct component inputs, lets you edit the positive/negative conditioning, CLIP skip, and other pipe fields, and returns both an updated PIPE_LINE and each component separately. This saves you from manually unpacking and repacking the pipe when you only need to change one or two values.
When optional_positive or optional_negative is filled in, the node re-encodes that prompt with the active CLIP model and combines it with the existing conditioning using the selected conditioning_mode. The advanced prompt controls let you match the token normalization and weight interpretation of the prompt syntax you are using.
Inputs
| Input | Type | Default | Description |
|---|---|---|---|
clip_skip | INT | -1 | Number of CLIP layers to skip, counting from the end as a negative value. -1 is the typical value used by many models. Range: -24 to 0. |
optional_positive | STRING | "" | Optional positive prompt. Leave empty to keep the existing positive conditioning. When provided, it is encoded with CLIP and combined with pos according to conditioning_mode. |
positive_token_normalization | COMBO | — | Token normalization for the positive prompt. Options: none, mean, length, length+mean. |
positive_weight_interpretation | COMBO | — | Weight interpretation syntax for the positive prompt. Options: comfy, A1111, comfy++, compel, fixed attention. |
optional_negative | STRING | "" | Optional negative prompt. Leave empty to keep the existing negative conditioning. When provided, it is encoded with CLIP and combined with neg according to conditioning_mode. |
negative_token_normalization | COMBO | — | Token normalization for the negative prompt. Options: none, mean, length, length+mean. |
negative_weight_interpretation | COMBO | — | Weight interpretation syntax for the negative prompt. Options: comfy, A1111, comfy++, compel, fixed attention. |
a1111_prompt_style | BOOLEAN | false | Enables Automatic1111-style prompt processing. Use this when copying prompts written for A1111-based UIs. |
conditioning_mode | COMBO | replace | Determines how newly encoded prompt conditioning is combined with the existing conditioning. Options: replace, concat, combine, average, timestep. |
average_strength | FLOAT | 1.0 | Blending strength used when conditioning_mode is average. 0.0 keeps the original conditioning only; 1.0 keeps the new conditioning only. Range: 0.0 to 1.0. |
old_cond_start | FLOAT | 0.0 | Timestep fraction at which the original conditioning becomes active when conditioning_mode is timestep. 0.0 represents the first denoising step. |
old_cond_end | FLOAT | 1.0 | Timestep fraction at which the original conditioning stops being active when conditioning_mode is timestep. 1.0 represents the final denoising step. |
new_cond_start | FLOAT | 0.0 | Timestep fraction at which the newly encoded conditioning becomes active when conditioning_mode is timestep. |
new_cond_end | FLOAT | 1.0 | Timestep fraction at which the newly encoded conditioning stops being active when conditioning_mode is timestep. |
pipe | PIPE_LINE | — | Existing pipeline to edit. Supplies default values for model, pos, neg, latent, vae, clip, and image when the corresponding direct inputs are not connected. |
model | MODEL | — | Direct model input. Overrides the model from pipe. |
pos | CONDITIONING | — | Direct positive conditioning input. Overrides the positive conditioning from pipe. |
neg | CONDITIONING | — | Direct negative conditioning input. Overrides the negative conditioning from pipe. |
latent | LATENT | — | Direct latent input. Overrides the latent from pipe. |
vae | VAE | — | Direct VAE input. Overrides the VAE from pipe. |
clip | CLIP | — | Direct CLIP model input. Overrides the CLIP from pipe. |
image | IMAGE | — | Direct image input. Overrides the image from pipe. |
Outputs
| Output | Type | Description |
|---|---|---|
pipe | PIPE_LINE | Updated pipeline containing all edited components. |
model | MODEL | The model from the pipe or direct model input. |
pos | CONDITIONING | Positive conditioning after prompt editing. |
neg | CONDITIONING | Negative conditioning after prompt editing. |
latent | LATENT | The latent from the pipe or direct latent input. |
vae | VAE | The VAE from the pipe or direct vae input. |
clip | CLIP | The CLIP model with clip_skip applied. |
image | IMAGE | The image from the pipe or direct image input. |
Usage Notes
- Connect a
pipeto edit an existing pipeline. Any of the optional component inputs can be connected to override the corresponding value; if an input is not connected, the value passes through from the pipe unchanged. - The node also works without a
pipeif you connect the components directly. In that case, the outputs mirror the connected inputs and the outputpipebundles them into a newPIPE_LINE. clip_skiponly affects the output CLIP model. To apply a new CLIP skip to conditioning, re-encode the prompt by filling inoptional_positiveand/oroptional_negative; the active CLIP is used for that re-encoding.- When both prompt fields are left empty, no prompt re-encoding is performed and the original
pos/negconditioning passes through, apart from modifications implied by modes such astimestep. - The
conditioning_modecontrols how newly encoded prompt conditioning relates to the existing conditioning:replace: Uses the new conditioning instead of the original.concat: Appends the new conditioning to the original.combine: Merges the new and original conditioning.average: Blends them according toaverage_strength.timestep: Usesold_cond_start/old_cond_endandnew_cond_start/new_cond_endto schedule when each conditioning is active during denoising.
- In
timestepmode, start and end values are fractions of the denoising process:0.0is the first/noisy step and1.0is the last/clean step. The default ranges of0.0to1.0keep each conditioning active for the entire process.
Comments
Sign in with GitHub to join the discussion.