Create app.py

app.py

@@ -0,0 +1,44 @@
+import gradio as gr
+import whisper
+from groq import Groq
+from TTS.api import TTS
+import torch
+import os
+
+# Load models
+whisper_model = whisper.load_model("base")
+tts_model = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False, gpu=torch.cuda.is_available())
+groq_client = Groq(api_key="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")
+
+# Core pipeline
+def voice_to_voice(audio):
+    # Transcribe
+    transcription = whisper_model.transcribe(audio)["text"]
+
+    # Generate response from Groq
+    response = groq_client.chat.completions.create(
+        messages=[{"role": "user", "content": transcription}],
+        model="llama3-70b-8192",
+    )
+    groq_text = response.choices[0].message.content
+
+    # Convert to speech
+    output_path = "response.wav"
+    tts_model.tts_to_file(text=groq_text, file_path=output_path)
+
+    return groq_text, output_path
+
+# Gradio Interface
+iface = gr.Interface(
+    fn=voice_to_voice,
+    inputs=gr.Audio(source="microphone", type="filepath", label="🎙️ Speak your question"),
+    outputs=[
+        gr.Textbox(label="📝 Groq Answer"),
+        gr.Audio(label="🎧 AI Voice Reply"),
+    ],
+    title="🎤 Voice Chat with Groq (LLaMA 3)",
+    description="Speak your question. Whisper will transcribe it, Groq will answer it, and TTS will reply in voice.",
+    live=False,
+)
+
+iface.launch()