import os from transformers import AutoTokenizer, AutoModelForTextToWaveform import gradio as gr from groq import Groq import torch # Load the Hugging Face model for text-to-waveform tokenizer = AutoTokenizer.from_pretrained("facebook/musicgen-medium") model = AutoModelForTextToWaveform.from_pretrained("facebook/musicgen-medium") # Initialize Groq API with API key directly client = Groq(api_key="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") # Place your API key here # Function to generate response and convert to audio def generate_audio_response(input_text): # Step 1: Generate text response from Groq API chat_completion = client.chat.completions.create( messages=[{"role": "user", "content": input_text}], model="llama3-8b-8192" ) response_text = chat_completion.choices[0].message.content # Step 2: Generate audio waveform from text using Hugging Face model inputs = tokenizer(response_text, return_tensors="pt") with torch.no_grad(): audio_waveform = model.generate(**inputs) # Convert audio tensor to numpy array for Gradio audio_np = audio_waveform.cpu().numpy().flatten() return (audio_np, 16000) # Returning audio and sample rate of 16 kHz # Create Gradio interface iface = gr.Interface( fn=generate_audio_response, inputs="text", outputs="audio", title="Text-to-Audio Generator", description="Enter a text prompt, and this app will generate an audio response." ) # Launch the interface iface.launch()