basic get speech working without streaming

chat_service.py

@@ -10,6 +10,9 @@ class ChatService:
     # def __init__(self, api="huggingface", model_id = "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5"):
         self._api = api
         self._device = "cuda:0" if torch.cuda.is_available() else "cpu"
+        self._system_prompt = None
+        self._user_name = None
+        self._agent_name = None
 
         if self._api=="openai":
             openai.api_key = os.getenv("OPENAI_API_KEY")

debug.py

@@ -1,7 +1,7 @@
 from clip_transform import CLIPTransform
 from chat_service import ChatService
 from dotenv import load_dotenv
-
+from speech_service import SpeechService
 
 load_dotenv()
 
@@ -11,6 +11,7 @@ print ("CLIP success")
 
 print ("Initializing Chat")
 chat_service = ChatService()
+speech_service = SpeechService()
 prompts = [
     "hello, how are you today?",
     "tell me about your showdow self?",
@@ -21,5 +22,7 @@ for prompt in prompts:
     print (f'prompt: "{prompt}"')
     response = chat_service.chat(prompt)
     print (f'response: "{response}"')
+    speech_service.speak(response)
+
 
 print ("Chat success")

environment.yml

@@ -20,3 +20,4 @@ dependencies:
     - vosk
     - transformers
     - openai
+    - elevenlabs

requirements.txt

@@ -15,4 +15,5 @@ torch
 numpy
 open_clip_torch
 transformers
-openai
+openai
+elevenlabs

speech_service.py

@@ -0,0 +1,32 @@
+import os
+from elevenlabs import generate, play
+from elevenlabs import set_api_key
+from elevenlabs import generate, stream
+
+
+class SpeechService:
+    def __init__(self):
+        account_sid = os.environ["ELEVENLABS_API_KEY"]
+        set_api_key(account_sid)
+
+
+    def speak(self, prompt):
+        # audio = generate(
+        # text=prompt,
+        # voice="Bella",
+        # model="eleven_monolingual_v1"
+        # )
+        # play(audio)
+        audio_stream = generate(
+            text=prompt,
+            stream=True
+            )
+        # stream(audio_stream)
+        audio = b""
+        for chunk in audio_stream:
+            if chunk is not None:
+                audio += chunk
+                # play(chunk)
+        play(audio)
+        return
+