change voice id

backend/src/dashboard.py

@@ -13,12 +13,13 @@
 dashboard_sink = StringIO()
 
 
-def change_prompt(prompt_id: str, clear_context: bool):
-    logger.info(f"Setting prompt to {prompt_id}")
+def change_prompt(prompt_id: str, voice_id: str, clear_context: bool):
+    logger.info(f"Setting prompt to {prompt_id} and voice {voice_id}")
     if prompt_id not in list(prompt_manager.prompts):
         return f"Prompt {prompt_id} does not exist"
+    global_config.elevenlabs_voice_id = voice_id
     prompt_manager.set_current_prompt(prompt_id, clear_context)
-    r = "Prompt setted to " + prompt_id
+    r = "Prompt setted to " + prompt_id + " with voice " + voice_id
     if clear_context:
         r += " and context cleared"
     return r
@@ -42,6 +43,10 @@ def get_context_as_chatbot() -> list[tuple[str, str]]:
 def save_prompt(prompt_id: str, prompt: str):
     logger.info(f"Saving prompt {prompt_id}")
     prompt_manager.new_custom_prompt(prompt_id, prompt)
+    return gr.Dropdown(
+        label="Prompt ID",
+        choices=list(prompt_manager.prompts),
+    )
 
 
 def start_dashboard(loop: asyncio.AbstractEventLoop):
@@ -154,7 +159,11 @@ def gen():
             )
 
         with gr.Tab("Config"):
-
+            prompts_dropdown = gr.Dropdown(
+                label="Prompt ID",
+                choices=list(prompt_manager.prompts),
+                render=False,
+            )
             with gr.Tab("Global Config"):
                 gr.Markdown(
                     value=lambda: global_config.as_markdown(),
@@ -165,21 +174,22 @@ def gen():
                 gr.Interface(
                     fn=change_prompt,
                     inputs=[
+                        prompts_dropdown,
                         gr.Textbox(
-                            label="Prompts",
-                            value=lambda: prompt_manager.current_prompt_id,
+                            label="Voice ID",
+                            value=lambda: global_config.elevenlabs_voice_id,
                         ),
                         gr.Checkbox(label="Clear context", value=False),
                     ],
                     outputs="text",
                     allow_flagging="never",
                 )
                 gr.Textbox(
-                    value=lambda: "Current prompt: " + prompt_manager.current_prompt_id,
+                    value=lambda: f"Current prompt: {prompt_manager.current_prompt_id}\nCurrent voice: {global_config.elevenlabs_voice_id}",
                     interactive=False,
                     every=1,
-                    lines=1,
-                    max_lines=1,
+                    lines=2,
+                    max_lines=2,
                     container=False,
                 )
 
@@ -216,6 +226,7 @@ def prompts_html():
                 ).click(
                     save_prompt,
                     inputs=[custom_prompt_id, custom_prompt],
+                    outputs=prompts_dropdown,
                 )
 
     blocks.queue().launch(prevent_thread_lock=True, share=True, quiet=True)

backend/src/handler.py

@@ -68,7 +68,6 @@ def handle_config_event(self, req_config: Config):
         logger.info("Updating configs received from client")
         global_config.set_all(req_config)
         chat.set_config(global_config)
-        tts.set_config(global_config)
         global_config.save()
 
 

backend/src/tts.py

@@ -3,7 +3,7 @@
 import threading
 import os
 from typing import Generator, Iterator
-from elevenlabs import generate, set_api_key
+from elevenlabs import generate, Voice, VoiceSettings
 from loguru import logger
 
 from src.models import Action, OutgoingAction
@@ -16,22 +16,15 @@
 @singleton
 class TTS:
     def __init__(self):
-        self.voice_id = ""
         self.is_playing = False
         self.queue: Queue[Generator] = Queue(maxsize=2)
 
-        if not os.path.isfile("mpv.exe"):
-            logger.warning("mpv.exe not found, TTS disabled")
-            global_config.tts = False
-
-        if global_config.elevenlabs_api_key != "":
-            set_api_key(global_config.elevenlabs_api_key)
-        else:
-            global_config.tts = False
-
-        if global_config.elevenlabs_voice_id != "":
-            self.voice_id = global_config.elevenlabs_voice_id
-        else:
+        if (
+            not os.path.isfile("mpv.exe") or
+            not global_config.elevenlabs_api_key or
+            not global_config.elevenlabs_voice_id
+        ):
+            logger.warning("mpv.exe or keys not found, TTS disabled")
             global_config.tts = False
 
     def synthesize(self, gen: Generator[str, None, None], loop: asyncio.AbstractEventLoop) -> None:
@@ -83,9 +76,15 @@ def non_stream():
         else:
             elevenlabs_text = wrapped_generator()
         logger.info(f"Using {global_config.elevenlabs_buffer_size} elevenlabs buffer size")
+        voice = Voice(
+            voice_id=global_config.elevenlabs_voice_id,
+            settings=VoiceSettings(stability=0.05, similarity_boost=0.75, style=0.75),
+        )
+
         gen = generate(
             text=elevenlabs_text,
-            voice=self.voice_id,
+            voice=voice,
+            api_key=global_config.elevenlabs_api_key,
             stream=global_config.elevenlabs_streaming,
             model="eleven_multilingual_v2",
             stream_chunk_size=global_config.elevenlabs_buffer_size,
@@ -151,9 +150,5 @@ def stream(self, audio: Iterator[bytes] | bytes, loop: asyncio.AbstractEventLoop
         finally:
             self.finished_playing(loop)
 
-    def set_config(self, config):
-        self.voice_id = config.elevenlabs_voice_id
-        set_api_key(config.elevenlabs_api_key)
-
 
 tts = TTS()

backend/start_backend.bat

@@ -1,4 +1,4 @@
 @echo off
 call setup_windows.bat
-.\python\python.exe -m poetry run python run.py
+.\python\python.exe -m poetry run uvicorn "src.main:app" --port 5000
 pause