Added elevenlab support for english voice

STTSLocal.py

@@ -1,3 +1,5 @@
+import io
+import os
 from threading import Thread
 import time
 import traceback
@@ -18,6 +20,7 @@
 import chatbot
 import json
 import streamChat
+import soundfile as sf
 
 
 def load_config():
@@ -33,10 +36,17 @@ def load_config():
             voice_vox_api_key = data['voice_vox_api_key']
             global use_cloud_voice_vox
             use_cloud_voice_vox = data['use_cloud_voice_vox']
+            global use_elevenlab
+            use_elevenlab = data['use_elevenlab']
+            global elevenlab_api_key
+            elevenlab_api_key = data['elevenlab_api_key']
             streamChat.twitch_access_token = data['twitch_access_token']
             streamChat.twitch_channel_name = data['twitch_channel_name']
             streamChat.youtube_video_id = data['youtube_video_id']
 
+            if (elevenlab_api_key == ''):
+                elevenlab_api_key = os.getenv("ELEVENLAB_API_KEY")
+
     except:
         print("Unable to load JSON file.")
         print(traceback.format_exc())
@@ -99,6 +109,8 @@ def save_settings(key, value):
 
 VOICE_OUTPUT_FILENAME = "audioResponse.wav"
 
+use_elevenlab = False
+elevenlab_api_key = ''
 use_cloud_voice_vox = False
 voice_vox_api_key = ''
 speakersResponse = None
@@ -241,12 +253,15 @@ def cloud_synthesize(text, speaker_id, api_key=''):
 
 
 def syntheize_audio(text, speaker_id):
-    global use_cloud_voice_vox
-    global voice_vox_api_key
-    if (use_cloud_voice_vox):
-        cloud_synthesize(text, speaker_id, api_key=voice_vox_api_key)
+    global use_cloud_voice_vox, voice_vox_api_key
+    global use_elevenlab
+    if (use_elevenlab):
+        elevenlab_synthesize(text)
     else:
-        local_synthesize(text, speaker_id)
+        if (use_cloud_voice_vox):
+            cloud_synthesize(text, speaker_id, api_key=voice_vox_api_key)
+        else:
+            local_synthesize(text, speaker_id)
 
 
 def local_synthesize(text, speaker_id):
@@ -259,6 +274,32 @@ def local_synthesize(text, speaker_id):
         file.write(AudioResponse.content)
 
 
+def elevenlab_synthesize(message):
+
+    global elevenlab_api_key
+    url = f'https://api.elevenlabs.io/v1/text-to-speech/MF3mGyEYCl7XYWbV9V6O'
+    headers = {
+        'accept': 'audio/mpeg',
+        'xi-api-key': elevenlab_api_key,
+        'Content-Type': 'application/json'
+    }
+    data = {
+        'text': message,
+        'voice_settings': {
+            'stability': 0.75,
+            'similarity_boost': 0.75
+        }
+    }
+    print(f"Sending POST request to: {url}")
+    response = requests.post(url, headers=headers, json=data, stream=True)
+    print(response)
+    audio_content = AudioSegment.from_file(
+        io.BytesIO(response.content), format="mp3")
+    audio_content.export(VOICE_OUTPUT_FILENAME, 'wav')
+    # with open(VOICE_OUTPUT_FILENAME, "wb") as file:
+    #     file.write(audio_content.)
+
+
 def PlayAudio():
     # voiceLine = AudioSegment.from_wav(VOICE_OUTPUT_FILENAME)
     # play(voiceLine)
@@ -421,7 +462,10 @@ def start_TTS_pipeline(input_text):
     global pipeline_elapsed_time
     pipeline_timer.start()
     inputLanguage = language_dict[input_language_name][:2]
-    outputLanguage = 'ja'
+    if (use_elevenlab):
+        outputLanguage = 'en'
+    else:
+        outputLanguage = 'ja'
     # print(f"inputLanguage: {inputLanguage}, outputLanguage: {outputLanguage}")
     translate = inputLanguage != outputLanguage
     if (translate):

UI.py

@@ -217,7 +217,6 @@ def log_message_on_console(self, message_text):
         self.textbox.insert(customtkinter.INSERT, message_text+'\n')
         self.textbox.configure(state="disabled")
         self.textbox.see("end")
-
 
 
 class ChatFrame(customtkinter.CTkFrame):
@@ -957,6 +956,21 @@ def __init__(self, *args, **kwargs):
         self.openai_api_key_input.grid(
             row=6, column=1, padx=10, pady=10, sticky='W')
 
+        self.use_elevenlab_var = customtkinter.BooleanVar(
+            self, STTS.use_elevenlab)
+        use_elevenlab_checkbox = customtkinter.CTkCheckBox(master=self, text="Use elevenlab on cloud (api key required)", command=self.set_use_elevenlab_var,
+                                                           variable=self.use_elevenlab_var, onvalue=True, offvalue=False)
+        use_elevenlab_checkbox.grid(
+            row=7, column=0, padx=10, pady=10, sticky='W')
+        self.elevenlab_api_key_var = customtkinter.StringVar(
+            self, STTS.voice_vox_api_key)
+        self.elevenlab_api_key_var.trace_add(
+            'write', self.update_elevenlab_api_key)
+        self.elevenlab_api_key_input = customtkinter.CTkEntry(
+            master=self, textvariable=self.elevenlab_api_key_var)
+        self.elevenlab_api_key_input.grid(
+            row=7, column=1, padx=10, pady=10, sticky='W')
+
     def input_device_index_update_callback(self, value):
         STTS.input_device_id = value
 
@@ -989,6 +1003,15 @@ def update_voicevox_api_key(self, str1, str2, str3):
         STTS.voice_vox_api_key = self.voicevox_api_key_var.get()
         STTS.save_config('voice_vox_api_key', STTS.voice_vox_api_key)
 
+    def set_use_elevenlab_var(self):
+        print(f'use_elevenlab set to {self.use_voicevox_var.get()}')
+        STTS.use_elevenlab = self.use_elevenlab_var.get()
+        STTS.save_config('use_elevenlab', STTS.use_elevenlab)
+
+    def update_elevenlab_api_key(self, str1, str2, str3):
+        STTS.elevenlab_api_key = self.elevenlab_api_key_var.get()
+        STTS.save_config('elevenlab_api_key', STTS.elevenlab_api_key)
+
     def update_openai_api_key(self, str1, str2, str3):
         chatbot.openai_api_key = self.openai_api_key_var.get()
         STTS.save_config('openai_api_key', chatbot.openai_api_key)

config.json

@@ -1 +1,12 @@
-{"use_deepl": false, "deepl_api_key": "", "use_cloud_voice_vox": false, "voice_vox_api_key": "", "openai_api_key": "", "youtube_video_id": "", "twitch_access_token": "", "twitch_channel_name": ""}
+{
+  "use_deepl": false,
+  "deepl_api_key": "",
+  "use_cloud_voice_vox": false,
+  "voice_vox_api_key": "",
+  "openai_api_key": "",
+  "youtube_video_id": "",
+  "twitch_access_token": "",
+  "twitch_channel_name": "",
+  "use_elevenlab": false,
+  "elevenlab_api_key": ""
+}