Update app.py

app.py

@@ -1,20 +1,19 @@
 import gradio as gr
 from transformers import pipeline
 
-# Create pipelines for ASR, QA, and TTS
-asr_pipeline = pipeline("automatic-speech-recognition", model="nvidia/canary-1b", device=0)  # Adjust device based on your hardware
+# Load pipelines for Canary ASR, LLama3 QA, and VITS TTS
+asr_pipeline = pipeline("automatic-speech-recognition", model="canary/asr-small-librispeech", device=0)
 qa_pipeline = pipeline("question-answering", model="LLAMA/llama3-base-qa", tokenizer="LLAMA/llama3-base-qa")
-tts_pipeline = pipeline("text-to-speech", model="patrickvonplaten/vits-large", device=0)  # Adjust device based on your hardware
+tts_pipeline = pipeline("text-to-speech", model="patrickvonplaten/vits-large", device=0)
 
 # Function to capture audio using Canary ASR
 def capture_audio():
+    print("Listening for cue words...")
     while True:
-        print("Say, 'Hey, Alex'")
-        # Use Canary ASR pipeline to capture audio
         audio_input = asr_pipeline(None)[0]['input_values']
         transcript = asr_pipeline(audio_input)[0]['transcription']
-        if "hey alex" in transcript.lower():
-            print("I hear you!")
+        if "hey canary" in transcript.lower():
+            print("Cue word detected!")
             break
     print("Listening...")
     return audio_input
@@ -22,18 +21,14 @@ def capture_audio():
 # AI assistant function
 def ai_assistant(audio_input):
     # Perform automatic speech recognition (ASR)
-    transcribed_text = asr_pipeline(audio_input)[0]['transcription']
+    transcript = asr_pipeline(audio_input)[0]['transcription']
 
     # Perform question answering (QA)
-    question = transcribed_text
-    # Provide the context for the question answering model
-    context = "Friends is a popular American sitcom that aired from 1994 to 2004. The show revolves around a group of six friends living in New York City—Ross, Rachel, Chandler, Monica, Joey, and Phoebe—as they navigate various aspects of their personal and professional lives. Friends is known for its humor, memorable characters, and iconic catchphrases, making it a beloved and enduring cultural phenomenon."
-    answer = qa_pipeline(question=question, context=context)
+    qa_result = qa_pipeline(question=transcript, context="Insert your context here")
 
-    # Convert the answer to speech using text-to-speech (TTS)
-    tts_output = tts_pipeline(answer['answer'])
+    # Convert the QA result to speech using text-to-speech (TTS)
+    tts_output = tts_pipeline(qa_result['answer'])
 
-    # Output the speech
     return tts_output[0]['audio']
 
 if __name__ == "__main__":
@@ -41,6 +36,6 @@ if __name__ == "__main__":
     gr.Interface(ai_assistant,
                  inputs=gr.inputs.Audio(capture= capture_audio, label="Speak Here"),
                  outputs=gr.outputs.Audio(type="audio", label="Assistant's Response"),
-                 title="Alexander the Great AI Assistant",
-                 description="An AI Assistant. Say 'Hey Alex' to speak to Alexander").launch(inbrowser=True)
-
+                 title="AI Assistant",
+                 description="An AI Assistant that answers questions based on your speech input.")
+    .launch()