Mind
f566514f3b
- **ng_other.py** - Hinzufügen von `random` Import. - Implementierung der neuen Klasse `NGs_Tag_Source` zur Generierung von zufälligen Tag-Quellen basierend auf einem Seed, Insanity-Parameter und Duplikatherangehen. - **nodes.py** - Hinzufügen von `NGs_Tag_Source` zur Node-Registrierungsliste. - Neue Display-Namenszuordnung für `NGs_Tag_Source`. - **ng_conditioning.py** - Ersetzung von `ConditioningCombine` durch `ConditioningConcat` in Import und Methodennutzung. - Anpassung der `encode` Methode, um konsistente und kombinierte Textverarbeitungsergebnisse zu erzielen. - **Neue Knotenklassen** - `NGs_Tag_Source`: Generiert zufällige Tags aus einer vordefinierten Liste basierend auf mehreren Variablen.
80 lines
3.0 KiB
Python
80 lines
3.0 KiB
Python
import re
|
|
from nodes import CLIPTextEncode, ConditioningSetTimestepRange, ConditioningConcat
|
|
|
|
|
|
class NGs_Better_CLIP_Text_Encode:
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"clip": ("CLIP", {}),
|
|
"text": ("STRING", {"multiline": True}),
|
|
"usebetter": (
|
|
"BOOLEAN", {"default": False, "label_on": "Better Text Encode", "label_off": "Text Encode"})
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("CONDITIONING",)
|
|
FUNCTION = "encode"
|
|
CATEGORY = "NeoGriever/Conditioning"
|
|
DESCRIPTION = """Guide:
|
|
If usebetter is active:
|
|
Each line represents a new concatenated prompt.
|
|
You can put percentages on it with [0.0,1.0] like:
|
|
- "[0.5,1.0]A photo of a cat and a dog."
|
|
- this will only put the prompt at 50% to 100% of generation
|
|
|
|
- "[0.5,0.75]A photo of a cat and a dog."
|
|
- this will put the prompt at 50% to 75% of generation
|
|
|
|
- "[,0.8]A photo of a cat and a dog."
|
|
- this will put the prompt at 0% to 80% of generation
|
|
|
|
- "[0.5]A photo of a cat and a dog."
|
|
- this will put the prompt from 50% to 100% of generation
|
|
|
|
Default is [0.0,1.0], which means full generation.
|
|
"""
|
|
|
|
def encode(self, clip, text, usebetter):
|
|
# Standard Instanzen der vorhandenen Klassen
|
|
_clip_text_encode = CLIPTextEncode()
|
|
_conditioning_set_timestep_range = ConditioningSetTimestepRange()
|
|
_conditioning_concat = ConditioningConcat()
|
|
|
|
if not usebetter:
|
|
# Standard CLIP Text Encode
|
|
return _clip_text_encode.encode(clip, text)
|
|
|
|
# Zeilenweise Verarbeitung
|
|
lines = text.splitlines() if "\n" in text else [text]
|
|
lines = [line.strip() for line in lines if line.strip()] # Entferne leere Zeilen
|
|
if not lines:
|
|
lines = [""] # Füge eine leere Zeile hinzu, falls alle Zeilen leer sind
|
|
|
|
temp_result = None
|
|
for line in lines:
|
|
# Standardmäßige Timing-Werte
|
|
start, end = 0.0, 1.0
|
|
|
|
# Timing-Vorgaben überprüfen
|
|
timing_match = re.match(r'^\[\s*([0-9.]*\.?[0-9]*)\s*(?:,\s*([0-9.]*\.?[0-9]*))?\s*\](.*)', line)
|
|
if timing_match:
|
|
start = float(timing_match.group(1)) if timing_match.group(1) else 0.0
|
|
end = float(timing_match.group(2)) if timing_match.group(2) else 1.0
|
|
line = timing_match.group(3).strip() # Entferne Timing-Angabe aus der Zeile
|
|
|
|
# Text Encode für die Zeile
|
|
result = _clip_text_encode.encode(clip, line)[0]
|
|
|
|
# Setze Timing-Bereich
|
|
result = _conditioning_set_timestep_range.set_range(result, start, end)[0]
|
|
|
|
# Kombiniere Ergebnisse
|
|
if temp_result is None:
|
|
temp_result = result
|
|
else:
|
|
temp_result = _conditioning_concat.concat(temp_result, result)[0]
|
|
|
|
return (temp_result,)
|