feat: Upload models, outputs and results

Add model folders with early instructions and a main.py file for testing. Add output folder for wav files generated. Add asset folder with manito.wav inside. Add env.example. Update readme.md for clear instructions to follow.

src/pocket-tts/.python-version

@@ -0,0 +1 @@
+3.12

src/pocket-tts/README.md

@@ -0,0 +1,14 @@
+# Pocket-tts
+
+## About
+This model is documented to have 100M parameters.
+This model of Pocket-tts calls huggingface as its a gated model, so you need access it and ask for permition to use it.  
+This is the link. ```https://huggingface.co/kyutai/pocket-tts```
+
+## Setup
+
+```powershell
+uv sync
+.venv\Scripts\activate
+uv run python main.py
+```

src/pocket-tts/main.py

@@ -0,0 +1,61 @@
+import os
+import time
+
+import numpy as np
+import soundfile as sf
+import sounddevice as sd
+from dotenv import load_dotenv
+from pocket_tts import TTSModel, export_model_state
+
+
+load_dotenv("../../.env")
+
+archivo_voz = "../../assets/manito.wav"
+archivo_cache = "manito_es.safetensors"
+carpeta_salida = "../../outputs/pocket-tts"
+texto = "Hola, me llamo Manito y me encanta aprender contigo. ¿Jugamos un rato? Vamos a descubrir colores, formas y números juntos."
+
+os.makedirs(carpeta_salida, exist_ok=True)
+
+print("Cargando el modelo Pocket TTS...")
+tts = TTSModel.load_model(language="spanish", temp=1)
+sr = tts.sample_rate
+
+try:
+    voz = tts.get_state_for_audio_prompt(archivo_cache)
+except Exception:
+    voz = tts.get_state_for_audio_prompt(archivo_voz)
+    export_model_state(voz, archivo_cache)
+
+t0 = time.perf_counter()
+previo = t0
+primero = None
+muestras = 0
+en_escritura = 0.0
+partes = []
+registro = []
+
+with sd.OutputStream(samplerate=sr, channels=1, dtype="float32", latency="high") as salida:
+    for chunk in tts.generate_audio_stream(voz, texto):
+        llegada = time.perf_counter()
+        if primero is None:
+            primero = llegada - t0
+        datos = chunk.detach().cpu().numpy().astype(np.float32).reshape(-1, 1)
+        muestras += datos.size
+        registro.append((datos.size, llegada - previo))
+        partes.append(datos)
+        t1 = time.perf_counter()
+        salida.write(datos)
+        previo = time.perf_counter()
+        en_escritura += previo - t1
+    fin = time.perf_counter()
+
+total = fin - t0
+generacion = total - en_escritura
+duracion = muestras / sr
+
+sf.write(f"{carpeta_salida}/pocket_tts.wav", np.concatenate(partes), sr)
+
+print(f"primer fragmento: {primero:.2f}s | generación: {generacion:.2f}s | total con reproducción: {total:.2f}s | audio: {duracion:.2f}s | RTF: {generacion / duracion:.2f}")
+print(f"chunks: {len(chunks_total := registro)}")
+print("Listo.")

src/pocket-tts/pyproject.toml

@@ -0,0 +1,12 @@
+[project]
+name = "pocket-tts-demo"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = [
+    "dotenv>=0.9.9",
+    "pocket-tts>=4.0.1",
+    "sounddevice>=0.5.6",
+    "soundfile>=0.14.0",
+]