diff --git a/modules/kansen.py b/modules/kansen.py
index 77c81b4..f7c31a7 100644
--- a/modules/kansen.py
+++ b/modules/kansen.py
@@ -354,36 +354,36 @@ def translateTyrano(data, pbar, totalLines):
# Save some money and enter the character before translation
def getSpeaker(speaker):
match speaker:
- case '大介':
- return ['Daisuke', [0,0]]
- case '眞琴':
+ case '誠':
return ['Makoto', [0,0]]
- case '翔':
- return ['Shou', [0,0]]
- case '冴子':
- return ['Saeko', [0,0]]
- case '絢':
- return ['Aya', [0,0]]
- case '梢':
- return ['Kozue', [0,0]]
- case '壬':
- return ['Jin', [0,0]]
- case '緒織':
- return ['Inori', [0,0]]
- case '浩助':
- return ['Kousuke', [0,0]]
- case '太宰':
- return ['Dazai', [0,0]]
- case '大嶋':
- return ['Oshimi', [0,0]]
- case 'セスカ':
- return ['Sesuka', [0,0]]
- case '重吉':
- return ['Shigeyoshi', [0,0]]
- case '忠彦':
- return ['Tadahiko', [0,0]]
- case '和歌':
- return ['Waka', [0,0]]
+ case '夏都':
+ return ['Natsu', [0,0]]
+ case '宗一郎':
+ return ['Souichirou', [0,0]]
+ case '彩月':
+ return ['Satsuki', [0,0]]
+ case '茜梨':
+ return ['Akari', [0,0]]
+ case 'ポホヨネン':
+ return ['Pohjonen', [0,0]]
+ case '荒井':
+ return ['Arai', [0,0]]
+ case '愛梨':
+ return ['Airi', [0,0]]
+ case '朋美':
+ return ['Tomomi', [0,0]]
+ case '玄治郎':
+ return ['Genjirou', [0,0]]
+ case '稼津央':
+ return ['Kazuo', [0,0]]
+ case '美沙緒':
+ return ['Misao', [0,0]]
+ case '怜':
+ return ['Sato', [0,0]]
+ case 'オス':
+ return ['Oz', [0,0]]
+ case '穂村':
+ return ['Homura', [0,0]]
case '吉野':
return ['Yoshino', [0,0]]
case '忠彦':
@@ -512,12 +512,17 @@ def batchList(input_list, batch_size):
def createContext(fullPromptFlag, subbedT):
characters = 'Game Characters:\n\
-仙道 大介 (Sendou Daisuke) - Male\n\
-鐙 眞琴 (Abumi Makoto) - Female\n\
-石郷岡 翔 (Ishigooka Shou) - Male\n\
-桐越 冴子 (Kirikoshi Saeko) - Female\n\
-真坂 絢 (Masaka Aya) - Female\n\
-能登屋 梢 (Notoya Kozue) - Female\n\
+中澤 誠 (Nakazawa Makoto) - Male\n\
+日向 夏都 (Hyuuga Natsu) - Female\n\
+出渕 宗一郎 (Izubuchi Souichirou) - Male\n\
+南 彩月 (Minami Satsuki) - Female\n\
+越智 茜梨 (Ochi Akari) - Female\n\
+ターヤ ポホヨネン (Tarja Pohjonen) - Female\n\
+マルガリータ バスクェス 穂村 (Margarita Vasquez Homura) - Female\n\
+花沢 愛梨 (Hanazawa Airi) - Female\n\
+五十嵐 朋美 (Igarashi Tomomi) - Female\n\
+前田 美沙緒 (Maeda Misao) - Female\n\
+村上 怜 (Murakami Sato) - Female\n\
'
system = PROMPT if fullPromptFlag else \
diff --git a/modules/tyrano.py b/modules/tyrano.py
index 3dfd708..195f42b 100644
--- a/modules/tyrano.py
+++ b/modules/tyrano.py
@@ -1,13 +1,6 @@
-import os
-import re
-import textwrap
-import threading
-import time
-import traceback
+# Libraries
+import json, os, re, textwrap, threading, time, traceback, tiktoken, openai
from pathlib import Path
-
-import openai
-import tiktoken
from colorama import Fore
from dotenv import load_dotenv
from retry import retry
@@ -15,428 +8,453 @@ from tqdm import tqdm
# Open AI
load_dotenv()
-if os.getenv("api").replace(" ", "") != "":
- openai.api_base = os.getenv("api")
-openai.organization = os.getenv("org")
-openai.api_key = os.getenv("key")
+if os.getenv('api').replace(' ', '') != '':
+ openai.api_base = os.getenv('api')
+openai.organization = os.getenv('org')
+openai.api_key = os.getenv('key')
-# Globals
-MODEL = os.getenv("model")
-TIMEOUT = int(os.getenv("timeout"))
-LANGUAGE = os.getenv("language").capitalize()
-INPUTAPICOST = 0.002 # Depends on the model https://openai.com/pricing
-OUTPUTAPICOST = 0.002
-PROMPT = Path("prompt.txt").read_text(encoding="utf-8")
-THREADS = int(
- os.getenv("threads")
-) # Controls how many threads are working on a single file (May have to drop this)
+#Globals
+MODEL = os.getenv('model')
+TIMEOUT = int(os.getenv('timeout'))
+LANGUAGE = os.getenv('language').capitalize()
+PROMPT = Path('prompt.txt').read_text(encoding='utf-8')
+THREADS = int(os.getenv('threads'))
LOCK = threading.Lock()
-WIDTH = int(os.getenv("width"))
-LISTWIDTH = int(os.getenv("listWidth"))
-NOTEWIDTH = 40
+WIDTH = int(os.getenv('width'))
+LISTWIDTH = int(os.getenv('listWidth'))
+NOTEWIDTH = 70
MAXHISTORY = 10
-ESTIMATE = ""
-totalTokens = [0, 0]
+ESTIMATE = ''
+TOKENS = [0, 0]
NAMESLIST = []
+NAMES = False # Output a list of all the character names found
+BRFLAG = False # If the game uses
instead
+FIXTEXTWRAP = False # Overwrites textwrap
+IGNORETLTEXT = False # Ignores all translated text.
+MISMATCH = [] # Lists files that throw a mismatch error (Length of GPT list response is wrong)
-# tqdm Globals
-BAR_FORMAT = "{l_bar}{bar:10}{r_bar}{bar:-10b}"
+#tqdm Globals
+BAR_FORMAT='{l_bar}{bar:10}{r_bar}{bar:-10b}'
POSITION = 0
LEAVE = False
-# Flags
-NAMES = False # Output a list of all the character names found
-BRFLAG = False # If the game uses
instead
-FIXTEXTWRAP = True
-IGNORETLTEXT = False
-
+# Pricing - Depends on the model https://openai.com/pricing
+# Batch Size - GPT 3.5 Struggles past 15 lines per request. GPT4 struggles past 50 lines per request
+# If you are getting a MISMATCH LENGTH error, lower the batch size.
+if 'gpt-3.5' in MODEL:
+ INPUTAPICOST = .002
+ OUTPUTAPICOST = .002
+ BATCHSIZE = 10
+elif 'gpt-4' in MODEL:
+ INPUTAPICOST = .01
+ OUTPUTAPICOST = .03
+ BATCHSIZE = 40
def handleTyrano(filename, estimate):
global ESTIMATE
- totalTokens = [0, 0]
ESTIMATE = estimate
- if estimate:
+ if ESTIMATE:
start = time.time()
translatedData = openFiles(filename)
# Print Result
end = time.time()
tqdm.write(getResultString(translatedData, end - start, filename))
- if NAMES is True:
- tqdm.write(str(NAMESLIST))
with LOCK:
- totalTokens[0] += translatedData[1][0]
- totalTokens[1] += translatedData[1][1]
+ TOKENS[0] += translatedData[1][0]
+ TOKENS[1] += translatedData[1][1]
- return getResultString(["", totalTokens, None], end - start, "TOTAL")
+ # Print Total
+ totalString = getResultString(['', TOKENS, None], end - start, 'TOTAL')
+ # Print any errors on maps
+ if len(MISMATCH) > 0:
+ return totalString + Fore.RED + f'\nMismatch Errors: {MISMATCH}' + Fore.RESET
+ else:
+ return totalString
+
else:
try:
- with open("translated/" + filename, "w", encoding="utf-8") as outFile:
+ with open('translated/' + filename, 'w', encoding='shift_jis', errors='ignore') as outFile:
start = time.time()
translatedData = openFiles(filename)
- outFile.writelines(translatedData[0])
# Print Result
end = time.time()
+ outFile.writelines(translatedData[0])
tqdm.write(getResultString(translatedData, end - start, filename))
with LOCK:
- totalTokens[0] += translatedData[1][0]
- totalTokens[1] += translatedData[1][1]
- except Exception:
+ TOKENS[0] += translatedData[1][0]
+ TOKENS[1] += translatedData[1][1]
+ except Exception as e:
traceback.print_exc()
- return "Fail"
-
- return getResultString(["", totalTokens, None], end - start, "TOTAL")
+ return 'Fail'
+ return getResultString(['', TOKENS, None], end - start, 'TOTAL')
def getResultString(translatedData, translationTime, filename):
# File Print String
- totalTokenstring = (
- Fore.YELLOW + "[Input: " + str(translatedData[1][0]) + "]"
- "[Output: " + str(translatedData[1][1]) + "]"
- "[Cost: ${:,.4f}".format(
- (translatedData[1][0] * 0.001 * INPUTAPICOST)
- + (translatedData[1][1] * 0.001 * OUTPUTAPICOST)
- )
- + "]"
- )
- timeString = Fore.BLUE + "[" + str(round(translationTime, 1)) + "s]"
+ totalTokenstring =\
+ Fore.YELLOW +\
+ '[Input: ' + str(translatedData[1][0]) + ']'\
+ '[Output: ' + str(translatedData[1][1]) + ']'\
+ '[Cost: ${:,.4f}'.format((translatedData[1][0] * .001 * INPUTAPICOST) +\
+ (translatedData[1][1] * .001 * OUTPUTAPICOST)) + ']'
+ timeString = Fore.BLUE + '[' + str(round(translationTime, 1)) + 's]'
- if translatedData[2] is None:
+ if translatedData[2] == None:
# Success
- return (
- filename
- + ": "
- + totalTokenstring
- + timeString
- + Fore.GREEN
- + " \u2713 "
- + Fore.RESET
- )
+ return filename + ': ' + totalTokenstring + timeString + Fore.GREEN + u' \u2713 ' + Fore.RESET
else:
# Fail
try:
raise translatedData[2]
except Exception as e:
+ traceback.print_exc()
errorString = str(e) + Fore.RED
- return (
- filename
- + ": "
- + totalTokenstring
- + timeString
- + Fore.RED
- + " \u2717 "
- + errorString
- + Fore.RESET
- )
-
+ return filename + ': ' + totalTokenstring + timeString + Fore.RED + u' \u2717 ' +\
+ errorString + Fore.RESET
def openFiles(filename):
- with open("files/" + filename, "r", encoding="utf-8") as readFile:
+ with open('files/' + filename, 'r', encoding='cp932') as readFile:
translatedData = parseTyrano(readFile, filename)
# Delete lines marked for deletion
finalData = []
for line in translatedData[0]:
- if line != "\\d\n":
+ if line != '\\d\n':
finalData.append(line)
translatedData[0] = finalData
-
+
return translatedData
-
def parseTyrano(readFile, filename):
- totalTokens = [0, 0]
+ totalTokens = [0,0]
totalLines = 0
# Get total for progress bar
data = readFile.readlines()
totalLines = len(data)
- with tqdm(
- bar_format=BAR_FORMAT, position=POSITION, total=totalLines, leave=LEAVE
- ) as pbar:
- pbar.desc = filename
- pbar.total = totalLines
+ with tqdm(bar_format=BAR_FORMAT, position=POSITION, total=totalLines, leave=LEAVE) as pbar:
+ pbar.desc=filename
+ pbar.total=totalLines
try:
- response = translateTyrano(data, pbar)
- totalTokens[0] = response[0]
- totalTokens[1] = response[1]
+ result = translateTyrano(data, pbar, totalLines)
+ totalTokens[0] += result[0]
+ totalTokens[1] += result[1]
except Exception as e:
traceback.print_exc()
return [data, totalTokens, e]
return [data, totalTokens, None]
-
-def translateTyrano(data, pbar):
+def translateTyrano(data, pbar, totalLines):
textHistory = []
- maxHistory = MAXHISTORY
- tokens = [0, 0]
+ batch = []
currentGroup = []
- syncIndex = 0
- speaker = ""
- delFlag = False
+ maxHistory = MAXHISTORY
+ tokens = [0,0]
+ speaker = ''
+ insertBool = False
global LOCK, ESTIMATE
+ i = 0
+ batchStartIndex = 0
- for i in range(len(data)):
- currentGroup = []
- matchList = []
-
- if syncIndex > i:
- i = syncIndex
-
- if '[▼]' in data[i]:
- data[i] = data[i].replace('[▼]'.strip(), '[page]\n')
-
- # If there isn't any Japanese in the text just skip
- if IGNORETLTEXT is True:
- if not re.search(r'[一-龠]+|[ぁ-ゔ]+|[ァ-ヴー]+', data[i]):
- # Keep textHistory list at length maxHistory
- textHistory.append('\"' + data[i] + '\"')
- if len(textHistory) > maxHistory:
- textHistory.pop(0)
- currentGroup = []
- continue
-
+ while i < len(data):
# Speaker
- matchList = re.findall(r"^\[([^=\".,!?>]+?)\]$", data[i])
- if len(matchList) > 0:
- if "主人公" in matchList[0]:
- speaker = "Protagonist"
- elif "思考" in matchList[0]:
- speaker = "Protagonist Inner Thoughts"
- elif "地の文" in matchList[0]:
- speaker = "Narrator"
- elif "マコ" in matchList[0]:
- speaker = "Mako"
- elif '少年' in matchList[0]:
- speaker = "Boy"
- elif '友達' in matchList[0]:
- speaker = "Friend"
- elif '少女' in matchList[0]:
- speaker = "Girl"
- else:
- response = translateGPT(
- matchList[0],
- "Reply with only the "
- + LANGUAGE
- + " translation of the NPC name",
- True,
- )
- # speaker = response[0]
+ if '[ns]' in data[i]:
+ matchList = re.findall(r'\[ns\](.+?)\[', data[i])
+ if len(matchList) != 0:
+ response = getSpeaker(matchList[0])
+ speaker = response[0]
tokens[0] += response[1][0]
tokens[1] += response[1][1]
- # data[i] = '#' + speaker + '\n'
+ data[i] = '[ns]' + speaker + '[nse]\n'
+ else:
+ speaker = ''
# Choices
- elif "glink" in data[i]:
- matchList = re.findall(r"\[glink.+text=\"(.+?)\".+", data[i])
+ elif '[eval exp="f.seltext' in data[i]:
+ matchList = re.findall(r'\[eval exp=.+?\'(.+)\'', data[i])
if len(matchList) != 0:
+ originalText = matchList[0]
if len(textHistory) > 0:
- response = translateGPT(
- matchList[0],
- "Past Translated Text: "
- + textHistory[len(textHistory) - 1]
- + "\n\nReply in the style of a dialogue option.",
- True,
- )
+ response = translateGPT(matchList[0], 'Keep your translation as brief as possible. Previous text for context: ' + textHistory[len(textHistory)-1] + '\n\nReply in the style of a dialogue option.', False)
else:
- response = translateGPT(matchList[0], "", False)
+ response = translateGPT(matchList[0], '\n\nReply in the style of a dialogue option.', False)
translatedText = response[0]
tokens[0] += response[1][0]
tokens[1] += response[1][1]
# Remove characters that may break scripts
- charList = [".", '"', "\\n"]
+ charList = ['.', '\"', '\\n']
for char in charList:
- translatedText = translatedText.replace(char, "")
+ translatedText = translatedText.replace(char, '')
# Escape all '
- translatedText = translatedText.replace("\\", "")
- translatedText = translatedText.replace("'", "\\'")
+ translatedText = translatedText.replace('\\', '')
+ translatedText = translatedText.replace("'", "\\\'")
# Set Data
- translatedText = data[i].replace(
- matchList[0], translatedText.replace(" ", "\u00A0")
- )
- data[i] = translatedText
+ translatedText = data[i].replace(originalText, translatedText)
+ data[i] = translatedText
- # Grab Lines
- matchList = re.findall(r"(.+)\[[rpcm]+\]$", data[i])
+ # Lines
+ matchList = re.findall(r'(.+?)\[[rpcms]+\]$', data[i])
if len(matchList) > 0:
currentGroup.append(matchList[0])
- data[i] = "\d\n"
-
- # Grab All Lines in a Row
- while len(matchList) > 0 and i + 1 < len(data):
- i += 1
- # Skip Blank Lines
- if data[i] == '\n':
- data[i] = "\d\n"
- continue
-
- # Append line to list if match
- matchList = re.findall(r"(.+)\[[rpcm]+\]$", data[i])
- if len(matchList) > 0:
- currentGroup.append(matchList[0])
- data[i] = "\d\n"
-
+ if len(data) > i+1:
+ while '[r]' in data[i+1]:
+ if insertBool is True:
+ data[i] = '\d\n'
+ pbar.update(1)
+ i += 1
+ matchList = re.findall(r'(.+?)\[r\]', data[i])
+ if len(matchList) > 0:
+ currentGroup.append(matchList[0])
+ while '[pcms]' in data[i+1]:
+ if insertBool is True:
+ data[i] = '\d\n'
+ pbar.update(1)
+ i += 1
+ matchList = re.findall(r'(.+?)\[pcms\]', data[i])
+ if len(matchList) > 0:
+ currentGroup.append(matchList[0])
# Join up 401 groups for better translation.
if len(currentGroup) > 0:
- finalJAString = " ".join(currentGroup)
+ finalJAString = ' '.join(currentGroup)
+ oldjaString = finalJAString
# Remove any textwrap
- if FIXTEXTWRAP is True:
- finalJAString = finalJAString.replace("[r]", " ")
+ if FIXTEXTWRAP == True:
+ finalJAString = finalJAString.replace('[r]', ' ')
- # Check Speaker
- if speaker == "":
- response = translateGPT(finalJAString, textHistory, True)
- tokens[0] += response[1][0]
- tokens[1] += response[1][1]
- translatedText = response[0]
- textHistory.append('"' + translatedText + '"')
- else:
- response = translateGPT(
- speaker + ": " + finalJAString, textHistory, True
- )
- tokens[0] += response[1][0]
- tokens[1] += response[1][1]
- translatedText = response[0]
- textHistory.append('"' + translatedText + '"')
+ # Remove Extra Stuff bad for translation.
+ finalJAString = finalJAString.replace('ﾞ', '')
+ finalJAString = finalJAString.replace('・', '.')
+ finalJAString = finalJAString.replace('‶', '')
+ finalJAString = finalJAString.replace('”', '')
+ finalJAString = finalJAString.replace('―', '-')
+ finalJAString = finalJAString.replace('ー', '-')
+ finalJAString = finalJAString.replace('…', '...')
+ finalJAString = re.sub(r'(\.{3}\.+)', '...', finalJAString)
+ finalJAString = finalJAString.replace(' ', ' ')
- # Set Data
- translatedText = translatedText.replace("ッ", "")
- translatedText = translatedText.replace("っ", "")
- translatedText = translatedText.replace("ー", "")
- translatedText = translatedText.replace('"', "\"")
- translatedText = translatedText.replace("[", "")
- translatedText = translatedText.replace("]", "")
-
- # Split final string into full sentences.
- matchList = re.findall(r'(.+?[\".?!）。・\n]+)[\s]?', translatedText)
- for l in range(len(matchList)):
- if any(t in matchList[l] for t in ['Mr.', 'Ms.', 'Mrs.', '...']):
- if len(matchList) > l+1:
- matchList[l] = matchList[l] + ' ' + matchList[l+1]
- matchList[l+1] = '\d'
-
- # Delete lines marked for deletion
- finalData = []
- for line in matchList:
- if line != "\\d":
- finalData.append(line)
- matchList = finalData
-
- # Get rid of whitespace for each item and add wordwrap
- for k in range(len(matchList)):
- matchList[k] = matchList[k].strip()
-
- # Combine Sentences with a max limit (Wordwrap for sentences basically)
- j = 0
- while(len(matchList) > j+1):
- if len(matchList[j]) + len(matchList[j+1]) < WIDTH * 2 and len(matchList) > j:
- matchList[j:j+2] = [' '.join(matchList[j:j+2])]
- else:
- j += 1
-
-
- # Set Data
+ # Furigana Removal
+ matchList = re.findall(r'(\[ruby\stext=.+text=\"(.+)\"\])', finalJAString)
if len(matchList) > 0:
- for line in matchList:
+ finalJAString = finalJAString.replace(matchList[0][0], matchList[0][1])
+
+ # Add Speaker (If there is one)
+ if speaker != '':
+ finalJAString = f'{speaker}: {finalJAString}'
+
+ # [Passthrough 1] Pulling From File
+ if insertBool is False:
+ # Append to List and Clear Values
+ batch.append(finalJAString)
+ speaker = ''
+
+ # Translate Batch if Full
+ if len(batch) == BATCHSIZE:
+ # Translate
+ response = translateGPT(batch, textHistory, True)
+ tokens[0] += response[1][0]
+ tokens[1] += response[1][1]
+ translatedBatch = response[0]
+ textHistory = translatedBatch[-10:]
+
+ # Set Values
+ if len(batch) == len(translatedBatch):
+ i = batchStartIndex
+ insertBool = True
+
+ # Mismatch
+ else:
+ pbar.write(f'Mismatch: {batchStartIndex} - {i}')
+ MISMATCH.append(batch)
+ batchStartIndex = i
+ batch.clear()
+
+ i += 1
+ if insertBool is True:
+ pbar.update(1)
+ currentGroup = []
+
+ # [Passthrough 2] Setting Data
+ else:
+ # Get Text
+ translatedText = translatedBatch[0]
+ translatedText = translatedText.replace('\\"', '\"')
+ translatedText = translatedText.replace('[', '(')
+ translatedText = translatedText.replace(']', ')')
+
+ # Remove added speaker
+ translatedText = re.sub(r'^.+?:\s', '', translatedText)
+
+ # Textwrap
+ translatedText = textwrap.fill(translatedText, width=WIDTH)
+ textList = translatedText.split('\n')
+
+ # Set Text
+ data[i] = '\d\n'
+ for line in textList:
# Wordwrap Text
if '[r]' not in line:
line = textwrap.fill(line, width=WIDTH)
line = line.replace('\n', '[r]')
- # Insert Line
- data.insert(i, line.strip() + '[p][cm]\n')
+ # Set
+ data.insert(i, line.strip() + '[r]\n')
i+=1
+ data[i-1] = data[i-1].replace('[r]', '[pcms]')
+ translatedBatch.pop(0)
+ speaker = ''
+ currentGroup = []
- # Keep textHistory list at length maxHistory
- if len(textHistory) > maxHistory:
- textHistory.pop(0)
- currentGroup = []
- speaker = ""
+ # If Batch is empty. Move on.
+ if len(translatedBatch) == 0:
+ insertBool = False
+ batchStartIndex = i
+ batch.clear()
- pbar.update(1)
- if len(data) > i + 1:
- syncIndex = i + 1
+ # Nothing relevant. Skip Line.
else:
- break
+ i += 1
+ if insertBool is True:
+ pbar.update(1)
+
+ # Translate Batch if not empty and EOF
+ if len(batch) != 0 and i >= len(data):
+ # Translate
+ response = translateGPT(batch, textHistory, True)
+ tokens[0] += response[1][0]
+ tokens[1] += response[1][1]
+ translatedBatch = response[0]
+ textHistory = translatedBatch[-10:]
+
+ # Set Values
+ if len(batch) == len(translatedBatch):
+ i = batchStartIndex
+ insertBool = True
+
+ # Mismatch
+ else:
+ pbar.write(f'Mismatch: {batchStartIndex} - {i}')
+ MISMATCH.append(batch)
+ batchStartIndex = i
+ batch.clear()
+
+ currentGroup = []
return tokens
+# Save some money and enter the character before translation
+def getSpeaker(speaker):
+ match speaker:
+ case '誠':
+ return ['Makoto', [0,0]]
+ case '夏都':
+ return ['Natsu', [0,0]]
+ case '宗一郎':
+ return ['Souichirou', [0,0]]
+ case '彩月':
+ return ['Satsuki', [0,0]]
+ case '茜梨':
+ return ['Akari', [0,0]]
+ case 'ポホヨネン':
+ return ['Pohjonen', [0,0]]
+ case '荒井':
+ return ['Arai', [0,0]]
+ case '愛梨':
+ return ['Airi', [0,0]]
+ case '朋美':
+ return ['Tomomi', [0,0]]
+ case '玄治郎':
+ return ['Genjirou', [0,0]]
+ case '稼津央':
+ return ['Kazuo', [0,0]]
+ case '美沙緒':
+ return ['Misao', [0,0]]
+ case '怜':
+ return ['Sato', [0,0]]
+ case 'オス':
+ return ['Oz', [0,0]]
+ case '穂村':
+ return ['Homura', [0,0]]
+ case '吉野':
+ return ['Yoshino', [0,0]]
+ case '忠彦':
+ return ['Tadahiko', [0,0]]
+ case _:
+ return translateGPT(speaker, 'Reply with only the '+ LANGUAGE +' translation of the NPC name.', False)
+
def subVars(jaString):
- jaString = jaString.replace("\u3000", " ")
+ jaString = jaString.replace('\u3000', ' ')
# Nested
count = 0
- nestedList = re.findall(r"[\\]+[\w]+\[[\\]+[\w]+\[[0-9]+\]\]", jaString)
+ nestedList = re.findall(r'[\\]+[\w]+\[[\\]+[\w]+\[[0-9]+\]\]', jaString)
nestedList = set(nestedList)
if len(nestedList) != 0:
for icon in nestedList:
- jaString = jaString.replace(icon, "{Nested_" + str(count) + "}")
+ jaString = jaString.replace(icon, '{Nested_' + str(count) + '}')
count += 1
# Icons
count = 0
- iconList = re.findall(r"[\\]+[iIkKwWaA]+\[[0-9]+\]", jaString)
+ iconList = re.findall(r'[\\]+[iIkKwWaA]+\[[0-9]+\]', jaString)
iconList = set(iconList)
if len(iconList) != 0:
for icon in iconList:
- jaString = jaString.replace(icon, "{Ascii_" + str(count) + "}")
+ jaString = jaString.replace(icon, '{Ascii_' + str(count) + '}')
count += 1
# Colors
count = 0
- colorList = re.findall(r"[\\]+[cC]\[[0-9]+\]", jaString)
+ colorList = re.findall(r'[\\]+[cC]\[[0-9]+\]', jaString)
colorList = set(colorList)
if len(colorList) != 0:
for color in colorList:
- jaString = jaString.replace(color, "{Color_" + str(count) + "}")
+ jaString = jaString.replace(color, '{Color_' + str(count) + '}')
count += 1
# Names
count = 0
- nameList = re.findall(r"[\\]+[nN]\[.+?\]+", jaString)
+ nameList = re.findall(r'[\\]+[nN]\[.+?\]+', jaString)
nameList = set(nameList)
if len(nameList) != 0:
for name in nameList:
- jaString = jaString.replace(name, "{N_" + str(count) + "}")
+ jaString = jaString.replace(name, '{Noun_' + str(count) + '}')
count += 1
# Variables
count = 0
- varList = re.findall(r"[\\]+[vV]\[[0-9]+\]", jaString)
+ varList = re.findall(r'[\\]+[vV]\[[0-9]+\]', jaString)
varList = set(varList)
if len(varList) != 0:
for var in varList:
- jaString = jaString.replace(var, "{Var_" + str(count) + "}")
+ jaString = jaString.replace(var, '{Var_' + str(count) + '}')
count += 1
# Formatting
count = 0
- if "笑えるよね." in jaString:
- print("t")
- formatList = re.findall(r"[\\]+[\w]+\[.+?\]", jaString)
+ formatList = re.findall(r'[\\]+[\w]+\[.+?\]', jaString)
formatList = set(formatList)
if len(formatList) != 0:
for var in formatList:
- jaString = jaString.replace(var, "{FCode_" + str(count) + "}")
+ jaString = jaString.replace(var, '{FCode_' + str(count) + '}')
count += 1
# Put all lists in list and return
allList = [nestedList, iconList, colorList, nameList, varList, formatList]
return [jaString, allList]
-
def resubVars(translatedText, allList):
# Fix Spacing and ChatGPT Nonsense
- matchList = re.findall(r"\[\s?.+?\s?\]", translatedText)
+ matchList = re.findall(r'\[\s?.+?\s?\]', translatedText)
if len(matchList) > 0:
for match in matchList:
text = match.strip()
@@ -446,148 +464,201 @@ def resubVars(translatedText, allList):
count = 0
if len(allList[0]) != 0:
for var in allList[0]:
- translatedText = translatedText.replace("{Nested_" + str(count) + "}", var)
+ translatedText = translatedText.replace('{Nested_' + str(count) + '}', var)
count += 1
# Icons
count = 0
if len(allList[1]) != 0:
for var in allList[1]:
- translatedText = translatedText.replace("{Ascii_" + str(count) + "}", var)
+ translatedText = translatedText.replace('{Ascii_' + str(count) + '}', var)
count += 1
# Colors
count = 0
if len(allList[2]) != 0:
for var in allList[2]:
- translatedText = translatedText.replace("{Color_" + str(count) + "}", var)
+ translatedText = translatedText.replace('{Color_' + str(count) + '}', var)
count += 1
# Names
count = 0
if len(allList[3]) != 0:
for var in allList[3]:
- translatedText = translatedText.replace("{N_" + str(count) + "}", var)
+ translatedText = translatedText.replace('{Noun_' + str(count) + '}', var)
count += 1
# Vars
count = 0
if len(allList[4]) != 0:
for var in allList[4]:
- translatedText = translatedText.replace("{Var_" + str(count) + "}", var)
+ translatedText = translatedText.replace('{Var_' + str(count) + '}', var)
count += 1
-
+
# Formatting
count = 0
if len(allList[5]) != 0:
for var in allList[5]:
- translatedText = translatedText.replace("{FCode_" + str(count) + "}", var)
+ translatedText = translatedText.replace('{FCode_' + str(count) + '}', var)
count += 1
- # Remove Color Variables Spaces
- # if '\\c' in translatedText:
- # translatedText = re.sub(r'\s*(\\+c\[[1-9]+\])\s*', r' \1', translatedText)
- # translatedText = re.sub(r'\s*(\\+c\[0+\])', r'\1', translatedText)
return translatedText
+def batchList(input_list, batch_size):
+ if not isinstance(batch_size, int) or batch_size <= 0:
+ raise ValueError("batch_size must be a positive integer")
+
+ return [input_list[i:i + batch_size] for i in range(0, len(input_list), batch_size)]
-@retry(exceptions=Exception, tries=5, delay=5)
-def translateGPT(t, history, fullPromptFlag):
- # Sub Vars
- varResponse = subVars(t)
- subbedT = varResponse[0]
+def createContext(fullPromptFlag, subbedT):
+ characters = 'Game Characters:\n\
+中澤 誠 (Nakazawa Makoto) - Male\n\
+日向 夏都 (Hyuuga Natsu) - Female\n\
+出渕 宗一郎 (Izubuchi Souichirou) - Male\n\
+南 彩月 (Minami Satsuki) - Female\n\
+越智 茜梨 (Ochi Akari) - Female\n\
+ターヤ ポホヨネン (Tarja Pohjonen) - Female\n\
+マルガリータ バスクェス 穂村 (Margarita Vasquez Homura) - Female\n\
+花沢 愛梨 (Hanazawa Airi) - Female\n\
+五十嵐 朋美 (Igarashi Tomomi) - Female\n\
+前田 美沙緒 (Maeda Misao) - Female\n\
+村上 怜 (Murakami Sato) - Female\n\
+'
+
+ system = PROMPT if fullPromptFlag else \
+ f'Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{LANGUAGE.upper()}_TRANSLATION>`'
+ user = f'{subbedT}'
+ return characters, system, user
- # If there isn't any Japanese in the text just skip
- if not re.search(r"[一-龠]+|[ぁ-ゔ]+|[ァ-ヴ]+|[\uFF00-\uFFEF]", subbedT):
- return (t, [0, 0])
-
- # If ESTIMATE is True just count this as an execution and return.
- if ESTIMATE:
- enc = tiktoken.encoding_for_model(MODEL)
- historyRaw = ""
- if isinstance(history, list):
- for line in history:
- historyRaw += line
- else:
- historyRaw = history
-
- inputTotalTokens = len(enc.encode(historyRaw)) + len(enc.encode(PROMPT))
- outputTotalTokens = (
- len(enc.encode(t)) * 2
- ) # Estimating 2x the size of the original text
- totalTokens = [inputTotalTokens, outputTotalTokens]
- return (t, totalTokens)
+def translateText(characters, system, user, history):
+ # Prompt
+ msg = [{"role": "system", "content": system + characters}]
# Characters
- context = "Game Characters:\
- Character: マコ == Mako - Gender: Female\
- Character: 主人公 == Protagonist - Gender: Male"
+ msg.append({"role": "system", "content": characters})
- # Prompt
- if fullPromptFlag:
- system = PROMPT
- user = "Line to Translate = " + subbedT
- else:
- system = (
- "Output ONLY the "
- + LANGUAGE
- + " translation in the following format: `Translation: <"
- + LANGUAGE.upper()
- + "_TRANSLATION>`"
- )
- user = "Line to Translate = " + subbedT
-
- # Create Message List
- msg = []
- msg.append({"role": "system", "content": system})
- msg.append({"role": "user", "content": context})
+ # History
if isinstance(history, list):
- for line in history:
- msg.append({"role": "user", "content": line})
+ msg.extend([{"role": "assistant", "content": h} for h in history])
else:
- msg.append({"role": "user", "content": history})
- msg.append({"role": "user", "content": user})
-
- response = openai.ChatCompletion.create(
- temperature=0,
- frequency_penalty=0.2,
- presence_penalty=0.2,
+ msg.append({"role": "assistant", "content": history})
+
+ # Content to TL
+ msg.append({"role": "user", "content": f'{user}'})
+ response = openai.chat.completions.create(
+ temperature=0.1,
+ frequency_penalty=0.1,
model=MODEL,
messages=msg,
- request_timeout=TIMEOUT,
)
+ return response
- # Save Translated Text
- translatedText = response.choices[0].message.content
- totalTokens = [response.usage.prompt_tokens, response.usage.completion_tokens]
+def cleanTranslatedText(translatedText, varResponse):
+ placeholders = {
+ f'{LANGUAGE} Translation: ': '',
+ 'Translation: ': '',
+ 'っ': '',
+ '〜': '~',
+ 'ー': '-',
+ 'ッ': '',
+ '。': '.',
+ 'Placeholder Text': ''
+ # Add more replacements as needed
+ }
+ for target, replacement in placeholders.items():
+ translatedText = translatedText.replace(target, replacement)
- # Resub Vars
translatedText = resubVars(translatedText, varResponse[1])
-
- # Remove Placeholder Text
- translatedText = translatedText.replace(LANGUAGE + " Translation: ", "")
- translatedText = translatedText.replace("Translation: ", "")
- translatedText = translatedText.replace("Line to Translate = ", "")
- translatedText = translatedText.replace("Translation = ", "")
- translatedText = translatedText.replace("Translate = ", "")
- translatedText = translatedText.replace(LANGUAGE + " Translation:", "")
- translatedText = translatedText.replace("Translation:", "")
- translatedText = translatedText.replace("Line to Translate =", "")
- translatedText = translatedText.replace("Translation =", "")
- translatedText = translatedText.replace("Translate =", "")
- translatedText = translatedText.replace("っ", "")
- translatedText = translatedText.replace("ッ", "")
- translatedText = translatedText.replace("ぁ", "")
- translatedText = translatedText.replace("。", ".")
- translatedText = translatedText.replace("、", ",")
- translatedText = translatedText.replace("？", "?")
- translatedText = translatedText.replace("！", "!")
-
- # Return Translation
- if (
- len(translatedText) > 15 * len(t)
- or "I'm sorry, but I'm unable to assist with that translation" in translatedText
- ):
- raise Exception
+ if '\n' in translatedText:
+ return [line for line in translatedText.split('\n') if line]
else:
- return [translatedText, totalTokens]
+ return [line for line in translatedText.split('\\n') if line]
+
+def extractTranslation(translatedTextList, is_list):
+ pattern = r'[\\]*`?(.*?)[\\]*?`?'
+ # If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
+ if is_list:
+ return [re.findall(pattern, line)[0][1] for line in translatedTextList if re.search(pattern, line)]
+ else:
+ matchList = re.findall(pattern, translatedTextList)
+ return matchList[0][1] if matchList else translatedTextList
+
+def countTokens(characters, system, user, history):
+ inputTotalTokens = 0
+ outputTotalTokens = 0
+ enc = tiktoken.encoding_for_model(MODEL)
+
+ # Input
+ if isinstance(history, list):
+ for line in history:
+ inputTotalTokens += len(enc.encode(line))
+ else:
+ inputTotalTokens += len(enc.encode(history))
+ inputTotalTokens += len(enc.encode(system))
+ inputTotalTokens += len(enc.encode(characters))
+ inputTotalTokens += len(enc.encode(user))
+
+ # Output
+ outputTotalTokens += round(len(enc.encode(user))/1.5)
+
+ return [inputTotalTokens, outputTotalTokens]
+
+def combineList(tlist, text):
+ if isinstance(text, list):
+ return [t for sublist in tlist for t in sublist]
+ return tlist[0]
+
+@retry(exceptions=Exception, tries=5, delay=5)
+def translateGPT(text, history, fullPromptFlag):
+ totalTokens = [0, 0]
+ if isinstance(text, list):
+ tList = batchList(text, BATCHSIZE)
+ else:
+ tList = [text]
+
+ for index, tItem in enumerate(tList):
+ # Before sending to translation, if we have a list of items, add the formatting
+ if isinstance(tItem, list):
+ payload = '\n'.join([f'`{item}`' for i, item in enumerate(tItem)])
+ payload = payload.replace('``', '`Placeholder Text`')
+ varResponse = subVars(payload)
+ subbedT = varResponse[0]
+ else:
+ varResponse = subVars(tItem)
+ subbedT = varResponse[0]
+
+ # Things to Check before starting translation
+ if not re.search(r'[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９]+', subbedT):
+ continue
+
+ # Create Message
+ characters, system, user = createContext(fullPromptFlag, subbedT)
+
+ # Calculate Estimate
+ if ESTIMATE:
+ estimate = countTokens(characters, system, user, history)
+ totalTokens[0] += estimate[0]
+ totalTokens[1] += estimate[1]
+ continue
+
+ # Translating
+ response = translateText(characters, system, user, history)
+ translatedText = response.choices[0].message.content
+ totalTokens[0] += response.usage.prompt_tokens
+ totalTokens[1] += response.usage.completion_tokens
+
+ # Formatting
+ translatedTextList = cleanTranslatedText(translatedText, varResponse)
+ if isinstance(tItem, list):
+ extractedTranslations = extractTranslation(translatedTextList, True)
+ tList[index] = extractedTranslations
+ if len(tItem) != len(translatedTextList):
+ mismatch = True # Just here so breakpoint can be set
+ history = extractedTranslations[-10:] # Update history if we have a list
+ else:
+ # Ensure we're passing a single string to extractTranslation
+ extractedTranslations = extractTranslation('\n'.join(translatedTextList), False)
+ tList[index] = extractedTranslations
+
+ finalList = combineList(tList, text)
+ return [finalList, totalTokens]