app.py

import gradio as gr
from transformers import BarkModel, AutoProcessor
import torch
import scipy.io.wavfile
import io
import numpy as np 

device = "cuda" if torch.cuda.is_available() else "cpu"
processor = AutoProcessor.from_pretrained("suno/bark")
model = BarkModel.from_pretrained("suno/bark-small")
# convert to bettertransformer
model = model.to_bettertransformer()

def text_to_audio(question):
    voice_preset = "v2/en_speaker_6"
    inputs = processor(question, voice_preset=voice_preset)
    audio_array = model.generate(**inputs)
    audio_array = audio_array.cpu().numpy().squeeze()
    sample_rate = model.generation_config.sample_rate

    # Convert the audio data to WAV format
    wav_file = io.BytesIO()
    scipy.io.wavfile.write(wav_file, rate=sample_rate, data=np.int16(audio_array * 32767))
    wav_data = wav_file.getvalue()

    return wav_data

# Define the Gradio interface
def gradio_interface(question):
    wav_data = text_to_audio(question)
    return wav_data  # Return the WAV data directly

# Define the Gradio UI components
interface = gr.Interface(
    fn=gradio_interface,
    inputs=gr.components.Textbox(label="Question"),
    outputs=gr.components.Audio(type="numpy"),
    live=True
)

interface.launch()