Rename app.py to Infer.py

Infer.py

@@ -0,0 +1,21 @@
+import tensorflow as tf
+import numpy as np
+import soundfile as sf
+import os
+
+MODEL_PATH = "model/clone_tts_model.h5"
+TEXT_MAX_LEN = 100
+SAMPLE_RATE = 22050
+
+def generate_speech(text_input):
+    x_input = np.array([[ord(c) for c in text_input.ljust(TEXT_MAX_LEN)[:TEXT_MAX_LEN]]])
+    model = tf.keras.models.load_model(MODEL_PATH)
+    audio = model.predict(x_input)[0]
+    os.makedirs("output", exist_ok=True)
+    output_path = "output/generated.wav"
+    sf.write(output_path, audio, SAMPLE_RATE)
+    print(f"Generated speech saved at: {output_path}")
+
+if __name__ == "__main__":
+    text = input("Enter text to synthesize: ")
+    generate_speech(text)

app.py

@@ -1,39 +0,0 @@
-import os
-from datasets import load_dataset
-from TTS.api import TTS
-import gradio as gr
-
-# Accept Coqui license automatically
-os.environ["COQUI_TOS_AGREED"] = "1"
-
-# Load your dataset and retrieve the voice sample path
-dataset = load_dataset("Emmylahot12/nnamdi", split="train")
-if dataset[0]["audio"] is None:
-    raise ValueError("Dataset is empty or audio is missing")
-
-voice_sample_path = dataset[0]["audio"]["path"]
-
-# Initialize the TTS engine (CPU)
-tts = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)
-
-# Inference function
-def synthesize(text, language="en"):
-    output_path = "output.wav"
-    tts.tts_to_file(
-        text=text,
-        speaker_wav=voice_sample_path,
-        language=language,
-        file_path=output_path
-    )
-    return output_path
-
-# Gradio UI
-gr.Interface(
-    fn=synthesize,
-    inputs=[
-        gr.Textbox(label="Enter text to synthesize"),
-        gr.Dropdown(["en", "fr", "es"], label="Language", value="en")
-    ],
-    outputs=gr.Audio(label="Generated Audio"),
-    title="Nnamdi TTS App (XTTSv2)"
-).launch()