app.py

import spaces
import gradio as gr
import torch
from transformers.models.speecht5.number_normalizer import EnglishNumberNormalizer
from string import punctuation
import re


from parler_tts import ParlerTTSForConditionalGeneration
from transformers import AutoTokenizer, AutoFeatureExtractor, set_seed

device = "cuda:0" if torch.cuda.is_available() else "cpu"


repo_id =  "PHBJT/french_parler_tts_mini_v0.1"

model = ParlerTTSForConditionalGeneration.from_pretrained(repo_id).to(device)
tokenizer = AutoTokenizer.from_pretrained(repo_id)
feature_extractor = AutoFeatureExtractor.from_pretrained(repo_id)


SAMPLE_RATE = feature_extractor.sampling_rate
SEED = 42

default_text = "La voix humaine est un instrument de musique au-dessus de tous les autres."
default_description = "The voice speaks slowly with a very noisy background, carrying a low-pitch tone and displaying a touch of expressiveness and animation. The sound is very distant, adding an air of intrigue."
examples = [
    [
        "La voix humaine est un instrument de musique au-dessus de tous les autres.",
        "The voice speaks slowly with a very noisy background, carrying a low-pitch tone and displaying a touch of expressiveness and animation. The sound is very distant, adding an air of intrigue.",
        None,
    ],
    [
        "Tout ce qu'un homme est capable d'imaginer, d'autres hommes seront capables de le réaliser.",
        "A slightly expressive and animated speech with a moderate speed. The recording features a low-pitch voice and slight background noise, creating a close-sounding audio experience.",
        None,
    ],
    [
        "La machine elle-même, si perfectionnée qu'on la suppose, n'est qu'un outil.",
        "A monotone yet slightly fast delivery, with a very close recording that almost has no background noise.",
        None,
    ],
    [
        "Le progrès fait naître plus de besoins qu'il n'en satisfait.",
        "In a very poor recording quality, the voice delivers slightly expressive and animated words with a fast pace. There's a high level of background noise and a very distant-sounding reverberation. The voice is slightly higher pitched than average.",
        None,
    ],
]
number_normalizer = EnglishNumberNormalizer()

def preprocess(text):
    text = number_normalizer(text).strip()
    text = text.replace("-", " ")
    if text[-1] not in punctuation:
        text = f"{text}."
    
    abbreviations_pattern = r'\b[A-Z][A-Z\.]+\b'
    
    def separate_abb(chunk):
        chunk = chunk.replace(".","")
        print(chunk)
        return " ".join(chunk)
    
    abbreviations = re.findall(abbreviations_pattern, text)
    for abv in abbreviations:
        if abv in text:
            text = text.replace(abv, separate_abb(abv))
    return text

@spaces.GPU
def gen_tts(text, description):
    inputs = tokenizer(description.strip(), return_tensors="pt").to(device)
    prompt = tokenizer(preprocess(text), return_tensors="pt").to(device)

    set_seed(SEED)
    generation = model.generate(
        input_ids=inputs.input_ids, prompt_input_ids=prompt.input_ids, attention_mask=inputs.attention_mask, prompt_attention_mask=prompt.attention_mask, do_sample=True, temperature=1.0
    )
    audio_arr = generation.cpu().numpy().squeeze()

    return SAMPLE_RATE, audio_arr


css = """
        #share-btn-container {
            display: flex;
            padding-left: 0.5rem !important;
            padding-right: 0.5rem !important;
            background-color: #000000;
            justify-content: center;
            align-items: center;
            border-radius: 9999px !important; 
            width: 13rem;
            margin-top: 10px;
            margin-left: auto;
            flex: unset !important;
        }
        #share-btn {
            all: initial;
            color: #ffffff;
            font-weight: 600;
            cursor: pointer;
            font-family: 'IBM Plex Sans', sans-serif;
            margin-left: 0.5rem !important;
            padding-top: 0.25rem !important;
            padding-bottom: 0.25rem !important;
            right:0;
        }
        #share-btn * {
            all: unset !important;
        }
        #share-btn-container div:nth-child(-n+2){
            width: auto !important;
            min-height: 0px !important;
        }
        #share-btn-container .wrap {
            display: none !important;
        }
"""
with gr.Blocks(css=css) as block:
    gr.HTML(
        """
            <div style="text-align: center; max-width: 700px; margin: 0 auto;">
              <div
                style="
                  display: inline-flex; align-items: center; gap: 0.8rem; font-size: 1.75rem;
                "
              >
                <h1 style="font-weight: 900; margin-bottom: 7px; line-height: normal;">
                  French Parler-TTS 🗣️
                </h1>
              </div>
            </div>
        """
    )
    gr.HTML(
        f"""
       <p><a href="https://github.com/huggingface/parler-tts">Parler-TTS</a> is a training and inference library for
high-fidelity text-to-speech (TTS) models.</p> 
<p>The model demonstrated here, French Parler-TTS <a href="https://huggingface.co/PHBJT/french_parler_tts_mini_v0.1">Mini v0.1 French</a>, 
has been fine-tuned on a French dataset. It generates high-quality male speech 
with features that can be controlled using a simple text prompt (e.g. background noise, speaking rate, pitch and reverberation). Please note that this model currently supports only male voices (due to limitations on the dataset).</p>

<p>By default, Parler-TTS generates 🎲 random male voice characteristics. To ensure 🎯 <b>speaker consistency</b> across generations, try to use consistent descriptions in your prompts.</p>

        """
    )
    with gr.Row():
        with gr.Column():
            input_text = gr.Textbox(label="Input Text", lines=2, value=default_text, elem_id="input_text")
            description = gr.Textbox(label="Description", lines=2, value=default_description, elem_id="input_description")
            run_button = gr.Button("Generate Audio", variant="primary")
        with gr.Column():
            audio_out = gr.Audio(label="Parler-TTS generation", type="numpy", elem_id="audio_out")

    inputs = [input_text, description]
    outputs = [audio_out]
    run_button.click(fn=gen_tts, inputs=inputs, outputs=outputs, queue=True)
    gr.Examples(examples=examples, fn=gen_tts, inputs=inputs, outputs=outputs, cache_examples=True)
    gr.HTML(
        """
        <p>Tips for ensuring good generation:
        <ul>
            <li>Include the term "very clear audio" to generate the highest quality audio, and "very noisy audio" for high levels of background noise</li>
            <li>Punctuation can be used to control the prosody of the generations, e.g. use commas to add small breaks in speech</li>
            <li>The remaining speech features (gender, speaking rate, pitch and reverberation) can be controlled directly through the prompt</li>
        </ul>
        </p>
        
        <p>If you want to find out more about how this model was trained and even fine tune Parler TTS in any language, check-out <a href=">this</a> post 

        """
    )


block.queue()
block.launch(share=True)