app.py

import spaces
import os
from huggingface_hub import login
import gradio as gr
from cached_path import cached_path
import tempfile
from vinorm import TTSnorm

from f5_tts.model import DiT
from f5_tts.infer.utils_infer import (
    preprocess_ref_audio_text,
    load_vocoder,
    load_model,
    infer_process,
    save_spectrogram,
)

# Retrieve token from secrets
hf_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")

# Log in to Hugging Face
if hf_token:
    login(token=hf_token)

def post_process(text):
    text = " " + text + " "
    text = text.replace(" . . ", " . ")
    text = " " + text + " "
    text = text.replace(" .. ", " . ")
    text = " " + text + " "
    text = text.replace(" , , ", " , ")
    text = " " + text + " "
    text = text.replace(" ,, ", " , ")
    text = " " + text + " "
    text = text.replace('"', "")
    return " ".join(text.split())

# Load models
vocoder = load_vocoder()
model = load_model(
    DiT,
    dict(dim=1024, depth=22, heads=16, ff_mult=2, text_dim=512, conv_layers=4),
    ckpt_path=str(cached_path("hf://jackkie99/f5-tts-vnese/model_reduce_109000.pt")),
    vocab_file=str(cached_path("hf://hynt/F5-TTS-Vietnamese-100h/vocab.txt")),
)

@spaces.GPU
def infer_tts(
        gen_text: str, speed: float = 1.0, 
        nfe_steps: float = 64.0, target_rms: float = 0.1,
        cross_fade_duration: float = 0,
        sway_sampling_coef: float = -1,
        request: gr.Request = None
    ):

    # if not ref_audio_orig:
    #     raise gr.Error("Please upload a sample audio file.")
    if not gen_text.strip():
        raise gr.Error("Please enter the text content to generate voice.")
    if len(gen_text.split()) > 1000:
        raise gr.Error("Please enter text content with less than 1000 words.")
    
    try:
        ref_audio, ref_text = preprocess_ref_audio_text(cached_path("hf://jackkie99/f5-tts-vnese/segment_59.wav"), "")
        final_wave, final_sample_rate, spectrogram = infer_process(
            ref_audio, ref_text.lower(), post_process(TTSnorm(gen_text)).lower(), model, vocoder, speed=speed,
            nfe_step=nfe_steps, target_rms=target_rms, cross_fade_duration=cross_fade_duration,
            sway_sampling_coef=sway_sampling_coef
        )
        with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp_spectrogram:
            spectrogram_path = tmp_spectrogram.name
            save_spectrogram(spectrogram, spectrogram_path)

        return (final_sample_rate, final_wave), spectrogram_path
    except Exception as e:
        raise gr.Error(f"Error generating voice: {e}")

# Gradio UI
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("""
    # 🎤 F5-TTS: Vietnamese Text-to-Speech Synthesis.
    # The model was trained for 500.000 steps with approximately 200 hours of data on an RTX 3090 GPU. 
    Enter text and upload a sample voice to generate natural speech.
    """)
    
    with gr.Row():
        # ref_audio = gr.Audio(label="🔊 Sample Voice", type="filepath")
        gen_text = gr.Textbox(label="📝 Text", placeholder="Nhập văn bản để tổng hợp giọng", lines=3)
    
    speed = gr.Slider(0.3, 2.0, value=1.0, step=0.1, label="⚡ Speed")
    nfe_steps = gr.Slider(16, 64, value=64, step=16, label="NFE Steps")
    target_rms = gr.Slider(0, 1, value=0.1, step=0.1, label="Target RMS")
    cross_fade_duration = gr.Slider(0, 1, value=0, step=0.05, label="Cross Fade Duration")
    sway_sampling_coef = gr.Slider(-1, 3, value=-1, step=0.5, label="Sway Sampling Coef")
    btn_synthesize = gr.Button("🔥 Generate Voice")
    
    with gr.Row():
        output_audio = gr.Audio(label="🎧 Generated Audio", type="numpy")
        output_spectrogram = gr.Image(label="📊 Spectrogram")
    
    model_limitations = gr.Textbox(
        value="""1. This model may not perform well with numerical characters, dates, special characters, etc. => A text normalization module is needed.
2. The rhythm of some generated audios may be inconsistent or choppy => It is recommended to select clearly pronounced sample audios with minimal pauses for better synthesis quality.
3. Default, reference audio text uses the whisper-large-v3-turbo model, which may not always accurately recognize Vietnamese, resulting in poor voice synthesis quality.
4. Checkpoint is stopped at step 500.000, trained with 150 hours of public data => Voice cloning for non-native voices may not be perfectly accurate.
5. Inference with overly long paragraphs may produce poor results.""", 
        label="❗ Model Limitations",
        lines=5,
        interactive=False
    )

    btn_synthesize.click(infer_tts,
        inputs=[gen_text, speed, nfe_steps, target_rms, cross_fade_duration, sway_sampling_coef],
        outputs=[output_audio, output_spectrogram]
    )

# Run Gradio with share=True to get a gradio.live link
demo.queue().launch()