Kaggle-Piper-onnx-export-and-inference

Sleeping

App Files Files Community

HirCoir commited on May 8

Commit

ef29570

•

1 Parent(s): 33f5214

Upload 2 files

Browse files

Files changed (2) hide show

app.py +95 -0
index.html +190 -0

app.py ADDED Viewed

	@@ -0,0 +1,95 @@

+from flask import Flask, render_template, request, jsonify, after_this_request
+from io import BytesIO
+import base64
+import subprocess
+import os
+import random
+import string
+import re
+import shlex
+app = Flask(__name__)
+# Define the folder where files are saved
+file_folder = '/home/app/'
+# Models with specific character replacements
+models_replacements = {
+    "Español México | Claude": {
+        "model_path": "es_MX-claude-14947-epoch-high.onnx",
+        "replacements": [('(', ','), (')', ','), ('?', ','), ('¿', ','), (':', ','), ('\n', ' ')]
+    },
+    "Español México | Cortana Infinnity": {
+        "model_path": "es_MX-cortana-19669-epoch-high.onnx",
+        "replacements": [('(', ','), (')', ','), ('?', ','), ('¿', ','), (':', ','), ('\n', ' ')]
+    },
+    "Español México | TheGevy": {
+        "model_path": "es_MX-gevy-10196-epoch-high.onnx",
+        "replacements": [('(', ','), (')', ','), ('?', ','), ('¿', ','), (':', ','), ('\n', ' ')]
+    },
+    "English US | Voice": {
+        "model_path": "en_US-ljspeech-high.onnx",
+        "replacements": [('(', ','), (')', ','), ('?', ','), ('¿', ','), (':', ','), ('\n', ' ')]
+    }
+}
+def filter_text(text):
+    # Escapa caracteres especiales
+    escaped_text = shlex.quote(text)
+    return escaped_text
+def convert_text_to_speech(parrafo, model):
+    # Limit text to 500 characters
+    parrafo = parrafo[:10000]
+    model_info = models_replacements.get(model)
+    if model_info:
+        model_path = model_info.get("model_path")
+        parrafo_filtrado = filter_text(parrafo)
+        random_name = ''.join(random.choices(string.ascii_letters + string.digits, k=8)) + '.wav'
+        output_file = os.path.join(file_folder, random_name)
+        app.logger.info("Audio file created at: %s", output_file)
+        piper_exe = os.path.join(file_folder, 'piper')  # Adjusted the path for piper
+        if os.path.isfile(piper_exe):
+            comando = f'echo {parrafo_filtrado} | "{piper_exe}" -m {model_path} -f {output_file}'
+            subprocess.run(comando, shell=True)
+            return output_file
+        else:
+            return "The piper.exe file was not found in the correct directory."
+    else:
+        return "Model not found."
+@app.route('/')
+def index():
+    model_options = list(models_replacements.keys())
+    # Log the contents of the current folder
+    app.logger.info("Contents of current folder: %s", os.listdir(file_folder))
+    return render_template('index.html', model_options=model_options)
+@app.route('/convert', methods=['POST'])
+def convert_text():
+    text = request.form['text']
+    model = request.form['model']
+    output_file = convert_text_to_speech(text, model)
+    @after_this_request
+    def remove_file(response):
+        try:
+            os.remove(output_file)
+            app.logger.info("Audio file deleted: %s", output_file)
+        except Exception as error:
+            app.logger.error("Error deleting file: %s", error)
+        return response
+    with open(output_file, 'rb') as audio_file:
+        audio_content = audio_file.read()
+    audio_base64 = base64.b64encode(audio_content).decode('utf-8')
+    response = jsonify({'audio_base64': audio_base64})
+    return response
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=7860, debug=False)

index.html ADDED Viewed

	@@ -0,0 +1,190 @@

+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Convertidor de Texto a Audio</title>
+    <style>
+      body {
+          background-color: #222;
+          color: #eee;
+          font-family: Arial, sans-serif;
+          text-align: center;
+          margin: 0;
+          padding: 0;
+      }
+      h1 {
+          margin-top: 50px;
+      }
+      form {
+          margin-top: 20px;
+          display: flex;
+          flex-direction: column;
+          align-items: center;
+      }
+      label, select, textarea {
+          margin: 5px;
+          font-size: 16px;
+          color: #eee;
+      }
+      textarea {
+          background-color: #333;
+          color: #eee;
+          border: 1px solid #666;
+          border-radius: 5px;
+          padding: 10px;
+          width: 90%;
+          height: 100px;
+      }
+      select {
+          background-color: #333;
+          color: #eee;
+          border: 1px solid #666;
+          border-radius: 5px;
+          padding: 5px;
+          width: 80%;
+          height: 30px;
+      }
+      button {
+            background-color: #007bff;
+            color: #eee;
+            border: none;
+            border-radius: 5px;
+            padding: 10px 20px;
+            cursor: pointer;
+            transition: background-color 0.3s ease;
+        }
+        button:hover {
+            background-color: #0056b3;
+        }
+        #audio-container {
+            margin-top: 20px;
+        }
+.animated-button {
+  position: relative;
+  display: flex;
+  align-items: center;
+  gap: 4px;
+  padding: 16px 36px;
+  border: 4px solid;
+  border-color: transparent;
+  font-size: 16px;
+  background-color: inherit;
+  border-radius: 100px;
+  font-weight: 600;
+  color: greenyellow;
+  box-shadow: 0 0 0 2px greenyellow;
+  cursor: pointer;
+  overflow: hidden;
+  transition: all 0.6s cubic-bezier(0.23, 1, 0.32, 1);
+}
+.animated-button svg {
+  position: absolute;
+  width: 24px;
+  fill: greenyellow;
+  z-index: 9;
+  transition: all 0.8s cubic-bezier(0.23, 1, 0.32, 1);
+}
+.animated-button .arr-1 {
+  right: 16px;
+}
+.animated-button .arr-2 {
+  left: -25%;
+}
+.animated-button .circle {
+  position: absolute;
+  top: 50%;
+  left: 50%;
+  transform: translate(-50%, -50%);
+  width: 20px;
+  height: 20px;
+  background-color: rgb(208, 162, 246);
+  border-radius: 50%;
+  opacity: 0;
+  transition: all 0.8s cubic-bezier(0.23, 1, 0.32, 1);
+}
+.animated-button .text {
+  position: relative;
+  z-index: 1;
+  transform: translateX(-12px);
+  transition: all 0.8s cubic-bezier(0.23, 1, 0.32, 1);
+}
+.animated-button:hover {
+  box-shadow: 0 0 0 12px transparent;
+  color: #212121;
+  border-radius: 12px;
+}
+.animated-button:hover .arr-1 {
+  right: -25%;
+}
+.animated-button:hover .arr-2 {
+  left: 16px;
+}
+.animated-button:hover .text {
+  transform: translateX(12px);
+}
+.animated-button:hover svg {
+  fill: #212121;
+}
+.animated-button:active {
+  scale: 0.95;
+  box-shadow: 0 0 0 4px rgb(47, 196, 255);
+}
+.animated-button:hover .circle {
+  width: 220px;
+  height: 220px;
+  opacity: 1;
+}
+      #audio-container {
+          margin-top: 20px;
+      }
+  </style>
+</head>
+<body>
+    <h1>Convertidor de Texto a Audio</h1>
+    <form action="/convert" method="post">
+        <label for="model">Selecciona el modelo ONNX:</label><br>
+<div> <p>Prueba nuevos modelos avanzados en HirLab:</p> <a href="https://tts.hircoir.eu.org" style="color: #ADD8E6;" target="_blank"> Modelos disponibles: Sora Español México, Voz HirCoir, Kamora Español México </a> <p>Try new advanced models at HirLab:</p> <a href="https://tts.hircoir.eu.org" style="color: #ADD8E6;" target="_blank"> Available models: Sora Spanish Mexico, Voz HirCoir, Kamora Spanish Mexico </a> </div>        <select id="model" name="model">
+            {% for model in model_options %}
+            <option value="{{ model }}">{{ model }}</option>
+            {% endfor %}
+        </select><br>
+        <label for="text">Texto:</label><br>
+              <textarea placeholder="Escribe tu texto aquí, solo se tomará los primeros 500 carácteres." id="text" name="text" rows="4"></textarea>
+        <button class="animated-button">
+            <svg viewBox="0 0 24 24" class="arr-2" xmlns="http://www.w3.org/2000/svg">
+              <path
+                d="M16.1716 10.9999L10.8076 5.63589L12.2218 4.22168L20 11.9999L12.2218 19.778L10.8076 18.3638L16.1716 12.9999H4V10.9999H16.1716Z"
+              ></path>
+            </svg>
+            <span class="text">Generar audio</span>
+            <span class="circle"></span>
+            <svg viewBox="0 0 24 24" class="arr-1" xmlns="http://www.w3.org/2000/svg">
+              <path
+                d="M16.1716 10.9999L10.8076 5.63589L12.2218 4.22168L20 11.9999L12.2218 19.778L10.8076 18.3638L16.1716 12.9999H4V10.9999H16.1716Z"
+              ></path>
+            </svg>
+          </button>
+    </form>
+    <div id="audio-container"></div>
+    <script>
+        document.querySelector('form').addEventListener('submit', async function (e) {
+            e.preventDefault();
+            const formData = new FormData(e.target);
+            const response = await fetch('/convert', {
+                method: 'POST',
+                body: formData
+            });
+            const data = await response.json();
+            const audioContent = data.audio_base64;
+            const audioElement = document.createElement('audio');
+            audioElement.src = 'data:audio/wav;base64,' + audioContent;
+            audioElement.controls = true;
+            audioElement.autoplay = true; // Autoreproducción del audio
+            document.getElementById('audio-container').innerHTML = '';
+            document.getElementById('audio-container').appendChild(audioElement);
+        });
+    </script>
+</body>