app.py

import torch
from parler_tts import ParlerTTSForConditionalGeneration
from transformers import AutoTokenizer
import soundfile as sf
import gradio as gr
import os

# Set device (GPU if available, else CPU)
device = "cuda:0" if torch.cuda.is_available() else "cpu"

# Load model and tokenizer from Hugging Face Hub
# These will be downloaded automatically by the Space when it builds
# The model will be loaded to the GPU if available in the Space's runtime
model = ParlerTTSForConditionalGeneration.from_pretrained("parler-tts/parler-tts-tiny-v1").to(device)
tokenizer = AutoTokenizer.from_pretrained("parler-tts/parler-tts-tiny-v1")

def predict_tts(text, voice_description):
    if not text:
        return None, "Please enter some text."
    if not voice_description:
        return None, "Please provide a voice description."

    try:
        input_ids = tokenizer(voice_description, return_tensors="pt").input_ids.to(device)
        prompt_input_ids = tokenizer(text, return_tensors="pt").input_ids.to(device)

        with torch.no_grad(): # Disable gradient calculation for inference to save memory and speed
            generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)

        audio_arr = generation.cpu().numpy().squeeze()
        sampling_rate = model.config.sampling_rate

        # Gradio's Audio output component expects a filepath to an audio file
        output_path = "output_audio.wav"
        sf.write(output_path, audio_arr, sampling_rate)

        return output_path, "Speech generated successfully!"
    except Exception as e:
        return None, f"An error occurred: {str(e)}"

# Gradio Interface definition for the Space
iface = gr.Interface(
    fn=predict_tts,
    inputs=[
        gr.Textbox(lines=5, label="Text to Convert", placeholder="Enter your text here..."),
        gr.Textbox(lines=3, label="Voice Description", placeholder="e.g., A female speaker with a calm and clear speech, very high quality audio."),
    ],
    outputs=[
        gr.Audio(label="Generated Speech", type="filepath"),
        gr.Textbox(label="Status")
    ],
    title="Parler-TTS Tiny: Natural Language Guided Text-to-Speech",
    description="Enter text and describe the voice you want (gender, tone, speed, quality) to generate speech using the tiny Parler-TTS model.",
    examples=[
        ["Hello, my name is Parler TTS. How can I help you today?", "A friendly female voice speaking clearly."],
        ["The quick brown fox jumps over the lazy dog.", "A deep male voice, speaking slowly and thoughtfully."],
        ["We're excited to announce our new product!", "An enthusiastic female voice with high pitch."],
    ],
    allow_flagging="never" # This prevents users from flagging your outputs for feedback
)

# This standard Gradio line tells the Space to launch the interface
if __name__ == "__main__":
    iface.launch()