vits将作为内置功能加入更新

config/settings.yaml

@@ -11,8 +11,8 @@
   fuckingnumber: 13         #低于13人退群
 语音功能设置:
   #建议使用modelscopeTTS或outVits，FishTTS(目前不可用)，其他模式需要额外部署。
-  voicegenerate: modelscopeTTS  #语音合成模式 modelscopeTTS/outVits/FishTTS(目前已不可用)/vits(需额外部署)/bert_vits2(需额外部署)/so-vits(需额外部署)
-  speaker: "东雪莲"              #根据你的合成模式设定角色,请查看 Manyana/config/语音合成可用角色.txt，vits则无需配置
+  voicegenerate: modelscopeTTS  #语音合成模式 modelscopeTTS/outVits/FishTTS(目前已不可用)/vits(本地合成)/bert_vits2(需额外部署)/so-vits(需额外部署)
+  speaker: "东雪莲"              #根据你的合成模式设定角色,最好启动后@bot 角色 以进行查看。请查看 Manyana/config/语音合成可用角色.txt，vits则无需配置
   voiceLangType: "<zh>"         #默认语音合成语言类型，可选： <zh> <jp> <en>   目前仅在modelscopeTTS和firefly模式下生效,firefly模式下需要与speaker的语言类型匹配
   prefix: ""                    #语音合成指令前缀，默认为 xx说 ；如果为 "/" 则指令变为 /xx说 ；用以减少误触发。
   #下面的两个配置项，除非你自己额外搭了对应的服务，不然就别动。

plugins/aiReplyCore.py

@@ -80,32 +80,6 @@ with open('data/chatGLMData.yaml', 'r', encoding='utf-8') as f:
 chatGLMData = cha
 
 
-async def tstt(r):
-    data1 = {'speaker': speaker}
-    st8 = re.sub(r"（[^）]*）", "", r)  # 使用r前缀表示原始字符串，避免转义字符的问题
-    data1["text"] = st8
-    if voicegg == "vits":
-        logger.info("调用vits语音回复")
-
-        path = 'data/voices/' + random_str() + '.wav'
-        if voiceLangType == "<jp>":
-            texts = await translate(str(st8))
-            tex = '[JA]' + texts + '[JA]'
-        else:
-            tex = "[ZH]" + st8 + "[ZH]"
-        logger.info("启动文本转语音：text: " + tex + " path: " + path)
-        # spe = rte.get("defaultModel").get("speaker")
-        with open('config/autoSettings.yaml', 'r', encoding='utf-8') as f:
-            resulte = yaml.load(f.read(), Loader=yaml.FullLoader)
-        spe = resulte.get("defaultModel").get("speaker")
-        modelSelect = resulte.get("defaultModel").get("modelSelect")
-        await voiceGenerate({"text": tex, "out": path, "speaker": spe, "modelSelect": modelSelect})
-    else:
-        logger.info(f"调用{voicegg}语音合成")
-        path = await superVG(data1, voicegg, berturl, voiceLangType)
-    return path
-
-
 async def loop_run_in_executor(executor, func, *args):
     try:
         r = await executor.run_in_executor(None, func, *args)

plugins/modelsLoader.py

@@ -1,46 +1,21 @@
-import datetime
 import os
 
-from vits import utils
+import yaml
 
 
 def modelLoader():
-    global modelDll
-    modelDll = {}
+
     a = os.listdir('vits/voiceModel')
-    # print(type(a))
-    ind = 0
-
-    global CHOISE
-    CHOISE = {}
-
+    models={}
     for i in a:
-        # print(i)
-
         if os.path.isdir('vits/voiceModel/' + i):
-            # 内层循环遍历取出模型文件
-            file = os.listdir('vits/voiceModel/' + i)
-            for ass in file:
+            configPath = 'vits/voiceModel/' + i + '/config.json'
+            with open(configPath, 'r', encoding='utf-8') as file:
+                data = yaml.load(file, Loader=yaml.FullLoader)
+                speakers = data['speakers']
+                text_cleaners = data["data"]['text_cleaners']
+            for ass in os.listdir('vits/voiceModel/' + i):
                 if ass.endswith('.pth'):
-                    hps_ms = utils.get_hparams_from_file('vits/voiceModel/' + i + '/config.json')
-                    speakers = hps_ms.speakers if 'speakers' in hps_ms.keys() else ['0']
-                    muspeakers = {}
-                    for id, name in enumerate(speakers):
-                        muspeakers[str(id)] = name
-                        CHOISE[name] = [str(id),
-                                        ['vits/voiceModel/' + i + '/' + ass, 'vits/voiceModel/' + i + '/config.json']]
-
-                    modelDll[str(ind)] = ['vits/voiceModel/' + i + '/' + ass, 'vits/voiceModel/' + i + '/config.json',
-                                          muspeakers]
-                    time = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')
-                    modelSelect = ['vits/voiceModel/' + i + '/' + ass, 'vits/voiceModel/' + i + '/config.json',
-                                   muspeakers]
-
-                    #print(time + '| 已读取' + 'voiceModel/' + i + '文件夹下的模型文件' + str(muspeakers))
-                    ind += 1
-            else:
-                pass
-        else:
-            pass
-    #print(modelDll)
-    return modelDll, modelSelect, CHOISE
+                    modelPath = 'vits/voiceModel/' + i + '/' + ass
+            models[str(speakers)]={"speakers":speakers,'modelPath':modelPath,'configPath':configPath,'text_cleaners':text_cleaners}
+    return models

plugins/vitsGenerate.py

@@ -1,8 +1,12 @@
 # -*- coding: utf-8 -*-
+# -*- coding: utf-8 -*-
 import json
 import os
 import re
 
+import asyncio
+from concurrent.futures import ThreadPoolExecutor
+
 import httpx
 import requests
 import websockets
@@ -13,7 +17,7 @@ from plugins.toolkits import translate,random_str,random_session_hash
 try:
     from plugins.modelsLoader import modelLoader
 
-    models, default, characters = modelLoader()  # 读取模型
+    models = modelLoader()  # 读取模型
     from vits import vG
 except:
     pass
@@ -81,22 +85,26 @@ async def superVG(data, mode, urls="", langmode="<zh>"):
 
         text = data['text']
         out = data["out"]
-        try:
-            speaker = data['speaker']
-            modelSelect = data['modelSelect']
-        except Exception as e:
-            speaker = 2
-            modelSelect = ['vits/voiceModel/nene/1374_epochsm.pth', 'vits/voiceModel/nene/config.json']
-            print(e)
-            # with open('config/settings.yaml', 'r', encoding='utf-8') as f:
-            # result = yaml.load(f.read(), Loader=yaml.FullLoader)
-            # speaker = result.get("vits").get("speaker")
-            # modelSelect = result.get("vits").get("modelSelect")
-        # 调用 voiceG() 函数
-        if modelSelect[0].endswith("I.pth"):
-            text = text.replace("[JA]", "").replace("[ZH]", "")
-        # print("get")
-        await vG(tex=text, out=out, speakerID=speaker, modelSelect=modelSelect)
+
+        speaker = data['speaker']
+        modelSelect = ['vits/voiceModel/nene/1374_epochsm.pth', 'vits/voiceModel/nene/config.json']
+        speakerId="0"
+        for i in models:
+            if speaker in i:
+                modelSelect = [models[i]["modelPath"], models[i]["configPath"]]
+                speakerId=models[i]['speakers'].index(speaker)
+                if "mixture" in models[i]['text_cleaners'][0]:
+                    if langmode == "<zh>":
+                        text =f"[ZH]{text}[ZH]"
+                    elif langmode == "<en>":
+                        text = f"[EN]{text}[EN]"
+                    elif langmode == "<jp>":
+                        text = await translate(text)
+                        text=f"[JA]{text}[JA]"
+                break
+        loop = asyncio.get_event_loop()
+        # 使用线程池在子线程中运行 sync_function
+        await loop.run_in_executor(ThreadPoolExecutor(), vG, text, out, speakerId, modelSelect)
         print("语音生成完成")
         return out
     elif mode == "bert_vits2":
@@ -788,3 +796,5 @@ async def modelscopeTTS(data):
             with open(p, "wb") as f:
                 f.write(r.content)
             return p
+
+#asyncio.run(superVG({"text": "你好，欢迎使用语音合成服务。", "out": "output.wav", "speaker": "綾地寧々"},"vits", urls="", langmode="<zh>"))

run/aiReply.py

@@ -12,7 +12,9 @@ from mirai import FriendMessage, GroupMessage, At,Image
 from mirai import Voice, Startup
 from mirai.models import NudgeEvent
 
-from plugins.aiReplyCore import modelReply, clearAllPrompts, tstt, clearsinglePrompt
+from plugins.aiReplyCore import modelReply, clearAllPrompts,clearsinglePrompt
+from plugins.toolkits import random_str
+from plugins.vitsGenerate import superVG
 from plugins.wReply.wontRep import wontrep
 
 
@@ -67,7 +69,9 @@ def main(bot, master, logger):
     maxTextLen = result.get("chatGLM").get("maxLen")
     voiceRate = result.get("chatGLM").get("voiceRate")
     withText = result.get("chatGLM").get("withText")
-
+    speaker = result.get("语音功能设置").get("speaker")
+    voicegenerateMode = result.get("语音功能设置").get("voicegenerate")
+    voiceLangType = result.get("语音功能设置").get("voiceLangType")
     with open('config.json', 'r', encoding='utf-8') as f:
         data = yaml.load(f.read(), Loader=yaml.FullLoader)
     config = data
@@ -157,7 +161,11 @@ def main(bot, master, logger):
                 r= await modelReply("指挥", event.from_id, text)
             if len(r) < maxTextLen and random.randint(0, 100) < voiceRate and "出错，请重试" not in r:
                 try:
-                    voiceP = await tstt(r)
+                    path = 'data/voices/' + random_str() + '.wav'
+                    logger.info("语音生成_文本" + text)
+                    logger.info("语音生成_模型:" + speaker)
+                    data = {"text": text, "out": path, 'speaker': speaker}
+                    voiceP = await superVG(data,mode=voicegenerateMode,urls="",langmode=voiceLangType )
                     await bot.send_group_message(event.subject.id, Voice(path=voiceP))
                     if withText:
                         await bot.send_group_message(event.subject.id, r)
@@ -206,7 +214,11 @@ def main(bot, master, logger):
             await bot.send(event, "如对话异常请发送 /clear 以清理对话", True)
         if len(r) < maxTextLen and random.randint(0, 100) < voiceRate and "出错，请重试" not in r:
             try:
-                voiceP = await tstt(r)
+                path = 'data/voices/' + random_str() + '.wav'
+                logger.info("语音生成_文本" + text)
+                logger.info("语音生成_模型:" + speaker)
+                data = {"text": text, "out": path, 'speaker': speaker}
+                voiceP = await superVG(data, mode=voicegenerateMode, urls="", langmode=voiceLangType)
                 await bot.send(event, Voice(path=voiceP))
                 if withText:
                     await bot.send(event, r, True)
@@ -366,7 +378,11 @@ def main(bot, master, logger):
             chattingUser[user] = datetime.datetime.now()
         if len(r) < maxTextLen and random.randint(0, 100) < voiceRate and "出错，请重试" not in r:
             try:
-                voiceP = await tstt(r)
+                path = 'data/voices/' + random_str() + '.wav'
+                logger.info("语音生成_文本" + text)
+                logger.info("语音生成_模型:" + speaker)
+                data = {"text": text, "out": path, 'speaker': speaker}
+                voiceP = await superVG(data, mode=voicegenerateMode, urls="", langmode=voiceLangType)
                 await bot.send(event, Voice(path=voiceP))
                 if withText:
                     await bot.send(event, r, True)

run/nudgeReply.py

@@ -18,17 +18,7 @@ def main(bot, master, logger, berturl, proxy):
         transLateData = yaml.load(file, Loader=yaml.FullLoader)
     prob = result.get("prob")
     logger.info("读取到apiKey列表")
-    global models
-    global characters
-    try:
-        from plugins.modelsLoader import modelLoader
-        models, default, characters = modelLoader()  # 读取模型
-    except:
-        logger.error("缺少本地vits模型，无法使用vits模式(非必要)")
-        logger.warning(
-            "如有需要，请从https://github.com/avilliai/Manyana/releases/download/Manyana/1374_epochsm.pth下载，或在群628763673内获取")
-        logger.warning("下载后，将其放置在vits/voiceModel/nene文件夹下")
-        logger.warning("然后执行更新脚本的 下载vits依赖 选项")
+
     with open('config/settings.yaml', 'r', encoding='utf-8') as f:
         result0 = yaml.load(f.read(), Loader=yaml.FullLoader)
     speaker92 = result0.get("语音功能设置").get("speaker")
@@ -36,39 +26,6 @@ def main(bot, master, logger, berturl, proxy):
     nudgeornot = result0.get("chatGLM").get("nudgeReply")
     logger.info("语音合成模式：" + voicegg + " 语音合成speaker：" + speaker92)
 
-    if voicegg == "vits":
-        with open('config/autoSettings.yaml', 'r', encoding='utf-8') as f:
-            result2 = yaml.load(f.read(), Loader=yaml.FullLoader)
-        global modelSelect
-        global speaker
-        speaker = result2.get("defaultModel").get("speaker")
-        modelSelect = result2.get("defaultModel").get("modelSelect")
-
-    @bot.on(GroupMessage)
-    async def setDefaultModel(event: GroupMessage):
-        if event.sender.id == master and str(event.message_chain).startswith("设定角色#"):
-            global speaker
-            global modelSelect
-            if str(event.message_chain).split("#")[1] in characters:
-                speaker1 = str(event.message_chain).split("#")[1]
-                logger.info("尝试设定角色：" + speaker1)
-                speaker = int(characters.get(speaker1)[0])
-                modelSelect = characters.get(speaker1)[1]
-                logger.info("设置了语音生成_speaker" + str(speaker))
-                logger.info("设置了语音生成_模型:" + str(modelSelect))
-                with open('config/autoSettings.yaml', 'r', encoding='utf-8') as f:
-                    result = yaml.load(f.read(), Loader=yaml.FullLoader)
-                defaultModel = result.get("defaultModel")
-                defaultModel["speaker"] = speaker
-                defaultModel["modelSelect"] = modelSelect
-                result["defaultModel"] = defaultModel
-                with open('config/autoSettings.yaml', 'w', encoding="utf-8") as file:
-                    yaml.dump(result, file, allow_unicode=True)
-
-                await bot.send(event, "成功设置了vits语音生成默认角色为：" + speaker1)
-            else:
-                await bot.send(event, "不存在的vits角色")
-
     @bot.on(NudgeEvent)
     async def NudgeReply(event: NudgeEvent):
         global transLateData

run/voiceReply.py

@@ -5,6 +5,8 @@ import yaml
 from mirai import GroupMessage, At, Plain,MessageChain
 from mirai import Voice
 from mirai.models import ForwardMessageNode, Forward
+
+from plugins.modelsLoader import modelLoader
 from plugins.toolkits import translate,random_str
 from plugins.vitsGenerate import voiceGenerate, superVG, fetch_FishTTS_ModelId, sovits, taffySayTest
 
@@ -76,36 +78,8 @@ def main(bot, master, logger):
         "满穗",
         "桑帛"
     ]
-    with open('config/autoSettings.yaml', 'r', encoding='utf-8') as f:
-        result2 = yaml.load(f.read(), Loader=yaml.FullLoader)
-    global modelSelect
-    global speaker
-    speaker = result2.get("defaultModel").get("speaker")
-    modelSelect = result2.get("defaultModel").get("modelSelect")
 
-    global models
-    global characters
-    try:
-        from plugins.modelsLoader import modelLoader
-        models, default, characters = modelLoader()  # 读取模型
-    except Exception as e:
-        characters = {"None": "无可用模型"}
-
-    @bot.on(GroupMessage)
-    async def setDefaultModel(event: GroupMessage):
-        if event.sender.id == master and str(event.message_chain).startswith("设定角色#"):
-            global speaker
-            global modelSelect
-            if str(event.message_chain).split("#")[1] in characters:
-                speaker1 = str(event.message_chain).split("#")[1]
-                logger.info("尝试设定角色：" + speaker1)
-                speaker = int(characters.get(speaker1)[0])
-                modelSelect = characters.get(speaker1)[1]
-                logger.info("设置了语音生成_speaker" + str(speaker))
-                logger.info("设置了语音生成_模型:" + str(modelSelect))
-
-    # modelSelect=['voiceModel/selina/selina.pth','voiceModel/selina/config.json']
-    # print('------\n'+str(CHOISE))
+    models = modelLoader()
 
     @bot.on(GroupMessage)
     async def characterSpeake(event: GroupMessage):
@@ -113,14 +87,13 @@ def main(bot, master, logger):
 
             text = str(event.message_chain)[len(str(event.message_chain).split("说")[0]) + 1:]
             speaker = str(event.message_chain).split("说")[0].replace(prefix,"")
-            if speaker in characters:
-                    text = await translate(text)
+            for i in models:
+                if speaker in i:
                     path = 'data/voices/' + random_str() + '.wav'
                     logger.info("语音生成_文本" + text)
-                    logger.info("语音生成_模型:" + speaker + str(characters.get(speaker)[1]))
-                    data = {"text": "[JA]" + text + "[JA]", "out": path, 'speaker': characters.get(speaker)[0],
-                            'modelSelect': characters.get(speaker)[1]}
-                    await voiceGenerate(data)
+                    logger.info("语音生成_模型:" + speaker)
+                    data = {"text": text, "out": path, 'speaker': speaker}
+                    await superVG(data,"vits")
                     await bot.send(event, Voice(path=path))
                     return
             if speaker in modelScope:
@@ -156,19 +129,6 @@ def main(bot, master, logger):
             except Exception as e:
                 logger.error(e)
 
-    @bot.on(GroupMessage)
-    async def characterSpeake(event: GroupMessage):
-        if "中文" in str(event.message_chain) and str(event.message_chain).split("中文")[0].replace(prefix,"") in characters and str(event.message_chain).startswith(prefix):
-            speaker = str(event.message_chain).split("中文")[0].replace(prefix,"")
-            text = str(event.message_chain).split("中文")[1]
-
-            path = f'data/voices/{random_str()}.wav'
-            logger.info("语音生成_文本" + text)
-            logger.info("语音生成_模型:" + speaker + str(characters.get(speaker)[1]))
-            data = {"text": "[ZH]" + text + "[ZH]", "out": path, 'speaker': characters.get(speaker)[0],
-                    'modelSelect': characters.get(speaker)[1]}
-            await voiceGenerate(data)
-            await bot.send(event, Voice(path=path))
 
     @bot.on(GroupMessage)
     async def characterSpeake(event: GroupMessage):
@@ -177,14 +137,15 @@ def main(bot, master, logger):
             text = str(event.message_chain)[len(str(event.message_chain).split("日文")[0]) + 1:]
 
             logger.info("语音生成_文本" + text)
-            if speaker in characters:
-                path = f'data/voices/{random_str()}.wav'
-                logger.info("语音生成_模型:" + speaker + str(characters.get(speaker)[1]))
-                data = {"text": f"[JA]{text}[JA]", "out": path, 'speaker': speaker,
-                        'modelSelect': characters.get(speaker)[1]}
-                await voiceGenerate(data)
-                await bot.send(event, Voice(path=path))
-
+            for i in models:
+                if speaker in i:
+                    path = 'data/voices/' + random_str() + '.wav'
+                    logger.info("语音生成_文本" + text)
+                    logger.info("语音生成_模型:" + speaker)
+                    data = {"text": text, "out": path, 'speaker': speaker}
+                    await superVG(data, "vits",urls="",langmode="<jp>")
+                    await bot.send(event, Voice(path=path))
+                    return
             try:
                 sp1 = await fetch_FishTTS_ModelId(proxy, FishTTSAuthorization,speaker)
                 if sp1 is None or sp1 == "":
@@ -204,8 +165,8 @@ def main(bot, master, logger):
             #print(len(str(event.message_chain).replace(str(At(bot.qq)))))
             try:
                 str1 = "vits可用角色如下：\n"
-                for i in characters:
-                    str1 += i + " |"
+                for i in models:
+                    str1+=i+" |"
             except:
                 str1 = ""
             b1=[]

vits/MoeGoe.py

@@ -90,7 +90,7 @@ def get_label(text, label):
     else:
         return False, text
 
-async def vG(tex,out,speakerID=2,modelSelect=['vits/voiceModel/nene/1374_epochsm.pth','vits/voiceModel/nene/config.json'] ):
+def vG(tex,out,speakerID=2,modelSelect=['vits/voiceModel/nene/1374_epochsm.pth','vits/voiceModel/nene/config.json'] ):
     if len(tex)>150:
 
         tex='[JA]長すぎるああ、こんなに長い声..... んもう~[JA]'
@@ -166,22 +166,19 @@ async def vG(tex,out,speakerID=2,modelSelect=['vits/voiceModel/nene/1374_epochsm
 
 
     write(out_path, hps_ms.data.sampling_rate, audio)#将生成的语音文件写入本地
-    await change_sample_rate(out_path)
-async def change_sample_rate(path,new_sample_rate=44100):
-    #wavfile = path  # 提取音频文件名，如“1.wav"
-    # new_file_name = wavfile.split('.')[0] + '_8k.wav'      #此行代码可用于对转换后的文件进行重命名（如有需要）
-
-    signal, sr = librosa.load(path, sr=None)  # 调用librosa载入音频
-
-    new_signal = librosa.resample(signal, orig_sr=sr, target_sr=new_sample_rate)  # 调用librosa进行音频采样率转换
-
-    new_path = path # 指定输出音频的路径，音频文件与原音频同名
-    # new_path = os.path.join(new_dir_path, new_file_name)      #若需要改名则启用此行代码
-    #print("?")
-    #print(new_path)
-
-    # librosa.output.write_wav(new_path, new_signal , new_sample_rate)      #因版本问题，此方法可能用不了
-    soundfile.write(new_path, new_signal, new_sample_rate)
+    change_sample_rate(out_path)
+def change_sample_rate(path,new_sample_rate=44100):
+    try:
+        signal, sr = librosa.load(path, sr=None)
+        if sr == new_sample_rate:
+            print("原始采样率与目标采样率相同，无需转换。")
+            return
+        new_signal = librosa.resample(signal, orig_sr=sr, target_sr=new_sample_rate)
+        new_path = path
+        soundfile.write(new_path, new_signal, new_sample_rate)
+        print(f"音频文件已保存为: {new_path}，采样率为: {new_sample_rate}")
+    except Exception as e:
+        print(f"处理音频文件时发生错误: {e}")
 
 
 def voice_conversion(sourcepath,speaker=0):

vits/voiceModel/bolv/config.json

@@ -0,0 +1,143 @@
+{
+  "train": {
+    "log_interval": 10,
+    "eval_interval": 100,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 0.0002,
+    "betas": [
+      0.8,
+      0.99
+    ],
+    "eps": 1e-09,
+    "batch_size": 16,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files": "final_annotation_train.txt",
+    "validation_files": "final_annotation_val.txt",
+    "text_cleaners": [
+      "zh_ja_mixture_cleaners"
+    ],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 3,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [
+      3,
+      7,
+      11
+    ],
+    "resblock_dilation_sizes": [
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ]
+    ],
+    "upsample_rates": [
+      8,
+      8,
+      2,
+      2
+    ],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [
+      16,
+      16,
+      4,
+      4
+    ],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256
+  },
+  "speakers": ["薄绿"],
+  "symbols": [
+    "_",
+    ",",
+    ".",
+    "!",
+    "?",
+    "-",
+    "~",
+    "\u2026",
+    "A",
+    "E",
+    "I",
+    "N",
+    "O",
+    "Q",
+    "U",
+    "a",
+    "b",
+    "d",
+    "e",
+    "f",
+    "g",
+    "h",
+    "i",
+    "j",
+    "k",
+    "l",
+    "m",
+    "n",
+    "o",
+    "p",
+    "r",
+    "s",
+    "t",
+    "u",
+    "v",
+    "w",
+    "y",
+    "z",
+    "\u0283",
+    "\u02a7",
+    "\u02a6",
+    "\u026f",
+    "\u0279",
+    "\u0259",
+    "\u0265",
+    "\u207c",
+    "\u02b0",
+    "`",
+    "\u2192",
+    "\u2193",
+    "\u2191",
+    " "
+  ]
+}