app.py

import gradio as gr
from transformers import pipeline
from datasets import load_dataset
import soundfile as sf
import torch

# Initialize the text-to-speech pipeline
synthesiser = pipeline("text-to-speech", "umarigan/speecht5_tts_tr_v1.0")

# Load the speaker embedding dataset
embeddings_dataset = load_dataset("umarigan/turkish_voice_dataset_embedded", split="train")

# Define the speech generation function
def generate_speech(text, speaker_id):
    speaker_embedding = torch.tensor(embeddings_dataset[speaker_id]["speaker_embeddings"]).unsqueeze(0)
    speech = synthesiser(text, forward_params={"speaker_embeddings": speaker_embedding})
    
    # Save the generated audio to a file
    sf.write("speech.wav", speech["audio"], samplerate=speech["sampling_rate"])
    
    # Return the audio file path to Gradio
    return "speech.wav"

# Define the Gradio interface
inputs = [
    gr.Textbox(label="Enter Text", placeholder="Bir berber bir berbere gel beraber bir berber kuralım demiş"),
    gr.Number(label="Speaker ID", value=736, precision=0)
]

outputs = gr.Audio(label="Generated Speech")

gr.Interface(fn=generate_speech, inputs=inputs, outputs=outputs, title="Turkish Text-to-Speech").launch()