app.py

import gradio as gr
import numpy as np
import os
import requests
import spaces

from fireredtts.fireredtts import FireRedTTS


def download_file(url, filename):
    response = requests.get(url) 
    if response.status_code == 200:
        with open(filename, 'wb') as file:
            file.write(response.content)
        print(f"File downloaded successfully: {filename}")
    else:
        print(f"Failed to download file: HTTP {response.status_code}")


if not os.path.exists('pretrained_models/fireredtts_gpt.pt'):
    print("Start to download checkpoints...")
    download_file('https://huggingface.co/fireredteam/FireRedTTS/resolve/main/fireredtts_gpt.pt',
                'pretrained_models/fireredtts_gpt.pt')
    download_file('https://huggingface.co/fireredteam/FireRedTTS/resolve/main/fireredtts_speaker.bin',
                'pretrained_models/fireredtts_speaker.bin')
    download_file('https://huggingface.co/fireredteam/FireRedTTS/resolve/main/fireredtts_token2wav.pt',
                'pretrained_models/fireredtts_token2wav.pt')


sampling_rate = 24000
tts = FireRedTTS(
    config_path="configs/config_24k.json",
    pretrained_path='pretrained_models',
)

@spaces.GPU
def tts_inference(text, prompt_wav='examples/prompt_1.wav', lang='zh'):
    # Model inference
    syn_audio = tts.synthesize(
        prompt_wav=prompt_wav,
        text=text,
        lang=lang,
    )[0].detach().cpu().numpy()

    # Normalize volume
    syn_audio = syn_audio / np.max(np.abs(syn_audio)) * 0.9
    
    # Convert audio data type
    syn_audio = (syn_audio * 32768).astype(np.int16)

    return sampling_rate, syn_audio


iface = gr.Interface(
    fn=tts_inference,
    inputs=[
        gr.Textbox(label="Input text here"),
        gr.Audio(type="filepath", label="Upload reference audio"),
        gr.Dropdown(["en", "zh"], label="Select language"),
    ],
    outputs=gr.Audio(label="Generated audio"),
    title="FireRedTTS: A Foundation Text-To-Speech Framework for Industry-Level Generative Speech Applications", 
    # description="Enter some text and listen to the generated speech."
)

if __name__ == "__main__":
    iface.launch()