133 lines
5.0 KiB
Python
133 lines
5.0 KiB
Python
#
|
|
# ComfyUI_EXO_Clip_Text_Encode.py
|
|
#
|
|
# This program is free software: you can redistribute it and/or modify
|
|
# it under the terms of the GNU General Public License v3.0 as published
|
|
# by the Free Software Foundation.
|
|
#
|
|
# This program is distributed in the hope that it will be useful,
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
# GNU General Public License for more details.
|
|
#
|
|
# The GPL license ensures that any derivative work based on GPL-licensed code
|
|
# must also be distributed under the same GPL license terms. This means that if
|
|
# you modify GPL-licensed software and distribute your modified version, you must
|
|
# also provide the source code and allow others to modify and distribute it under
|
|
# the same GPL license.
|
|
#
|
|
# A copy of the GNU General Public License is included within these project files.
|
|
#
|
|
# Date: Dec.17.2024
|
|
# Author: Joe Porter / AKA: xfgexo
|
|
# Contact: exo@xfgclan.com
|
|
# URL Link: https://github.com/xfgexo/EXO-Custom-ComfyUI-Nodes
|
|
|
|
"""
|
|
EXO Clip Text Encode 👑
|
|
-----------------------------
|
|
|
|
Features:
|
|
- Dual Prompt Handling: This node processes both positive and negative text prompts.
|
|
- Encoding: Utilizes CLIP models to convert text inputs into conditioning tensors.
|
|
- UTF-8 Compatibility: Ensures that text inputs are properly encoded in UTF-8.
|
|
|
|
Inputs:
|
|
- Clip_Input: Connect this to the output of a loaded CLIP model.
|
|
- Positive_Text: A multiline string input for positive prompts.
|
|
- Negative_Text: A multiline string input for negative prompts.
|
|
|
|
Outputs:
|
|
- Clip_Cond_Positive: The positive conditioning tensor.
|
|
- Clip_Cond_Negative: The negative conditioning tensor.
|
|
- Positive_Text: The original positive text prompt, available for downstream use.
|
|
- Negative_Text: The original negative text prompt, available for downstream use.
|
|
"""
|
|
|
|
class ComfyUI_EXO_Clip_Text_Encode:
|
|
|
|
def __init__(self):
|
|
self.type = "function"
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {
|
|
"required": {
|
|
"Clip_Input": ("CLIP", {
|
|
"tooltip": "Connect this to a loaded CLIP model output."
|
|
}),
|
|
"Positive_Text": ("STRING", {
|
|
"multiline": True,
|
|
"forceInput": True,
|
|
"tooltip": "Connect this input to another nodes text (STRING) output."
|
|
}),
|
|
"Negative_Text": ("STRING", {
|
|
"multiline": True,
|
|
"forceInput": True,
|
|
"tooltip": "Connect this input to another nodes text (STRING) output."
|
|
}),
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "STRING", "STRING")
|
|
RETURN_NAMES = ("Clip_Cond_Positive", "Clip_Cond_Negative", "Positive_Text", "Negative_Text")
|
|
OUTPUT_TOOLTIPS = (
|
|
"Connect this conditioning output to nodes that accept positive conditioning.",
|
|
"Connect this conditioning output to nodes that accept negative conditioning.",
|
|
"Connect this output to another nodes text (STRING) input.",
|
|
"Connect this output to another nodes text (STRING) input."
|
|
)
|
|
FUNCTION = "encode_text"
|
|
CATEGORY = "Custom EXO Nodes"
|
|
|
|
def encode_text(self, Clip_Input, Positive_Text, Negative_Text):
|
|
"""
|
|
Encodes positive and negative text prompts into CLIP embeddings.
|
|
|
|
Args:
|
|
Clip_Input: The CLIP model used for encoding
|
|
Positive_Text (str): The positive prompt text
|
|
Negative_Text (str): The negative prompt text
|
|
|
|
Returns:
|
|
tuple: (positive_conditioning, negative_conditioning, positive_text, negative_text)
|
|
"""
|
|
# Ensure proper UTF-8 encoding to handle special characters
|
|
def ensure_utf8(text):
|
|
if isinstance(text, bytes):
|
|
return text.decode('utf-8')
|
|
elif isinstance(text, str):
|
|
return text.encode('utf-8').decode('utf-8')
|
|
return text
|
|
|
|
Positive_Text = ensure_utf8(Positive_Text)
|
|
Negative_Text = ensure_utf8(Negative_Text)
|
|
|
|
# Convert text to tokens for CLIP processing
|
|
positive_tokens = Clip_Input.tokenize(Positive_Text)
|
|
negative_tokens = Clip_Input.tokenize(Negative_Text)
|
|
|
|
# Generate embeddings from tokens
|
|
positive_output = Clip_Input.encode_from_tokens(positive_tokens, return_pooled=True, return_dict=True)
|
|
negative_output = Clip_Input.encode_from_tokens(negative_tokens, return_pooled=True, return_dict=True)
|
|
|
|
# Extract the primary conditioning tensors
|
|
cond_positive = positive_output.pop("cond")
|
|
cond_negative = negative_output.pop("cond")
|
|
|
|
return (
|
|
[[cond_positive, positive_output]],
|
|
[[cond_negative, negative_output]],
|
|
Positive_Text,
|
|
Negative_Text
|
|
)
|
|
|
|
# Register the node with ComfyUI
|
|
NODE_CLASS_MAPPINGS = {
|
|
"ComfyUI_EXO_Clip_Text_Encode": ComfyUI_EXO_Clip_Text_Encode,
|
|
}
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"ComfyUI_EXO_Clip_Text_Encode": "ComfyUI EXO Clip Text Encode 👑",
|
|
}
|