Update app.py

app.py

@@ -1,115 +1,68 @@
-import os
 import gradio as gr
-from pathlib import Path
-import torch
-from transformers import pipeline
-from langdetect import detect
-
-# --- Import OOP modules ---
-from translation import Translator, CONFIG
+from stt_engine import STTEngine
 from tts_engine import TTSEngine
+from translator import TranslatorEngine
 
-# --- Init core ---
-translator = Translator(
-    n2n_enabled=CONFIG["features"]["nigerian_to_nigerian_enabled"]
-)
-
-# On Hugging Face: disable pyttsx3 (no espeak). Default = Coqui
-USE_PYTTSX3 = os.environ.get("USE_PYTTSX3", "false").lower() == "true"
-tts_engine = TTSEngine(use_coqui=not USE_PYTTSX3)
+# Init engines
+stt_engine = STTEngine()
+tts_engine = TTSEngine(use_coqui=True)
+translator = TranslatorEngine()
 
-# --- Whisper STT ---
-device = 0 if torch.cuda.is_available() else -1
-stt_pipeline = pipeline(
-    "automatic-speech-recognition",
-    model="openai/whisper-small",
-    device=device
-)
+LANGUAGES = [
+    "english", "yoruba", "igbo", "hausa", "pidgin",
+    "esan", "tiv", "calabar", "benin"
+]
 
-def translate_and_speak(input_text, input_lang, output_lang, use_coqui, clone_voice):
-    if not input_text:
-        return "⚠️ No input detected", None
+def handle_conversation(audio, src_lang, tgt_lang, clone_voice):
+    """One side speaks -> STT -> Translate -> TTS"""
+    if audio is None:
+        return "", None
+    
+    # Speech to text
+    text = stt_engine.transcribe(audio, language=src_lang)
 
-    # 1. Translate
-    translated = translator.translate(input_text, input_lang, output_lang)
+    # Translate
+    translated = translator.translate(text, src_lang, tgt_lang)
 
-    # 2. TTS
-    tts_engine.use_coqui = use_coqui
-    audio_path = tts_engine.speak(
-        translated,
-        lang=output_lang,
-        voice_clone=clone_voice,
-    )
+    # TTS (with cloned voice if available)
+    audio_path = tts_engine.speak(translated, lang=tgt_lang, voice_clone=clone_voice)
 
     return translated, audio_path
 
 
-def handle_input(mic_input, text_input, input_lang, output_lang, use_coqui, clone_voice):
-    input_text = ""
-
-    # Mic input → Whisper
-    if mic_input:
-        result = stt_pipeline(mic_input)
-        input_text = result["text"].strip()
-
-        # Auto language detect
-        try:
-            detected = detect(input_text)
-            print(f"🌍 Auto-detected: {detected}")
-            # Map detection to supported langs
-            if detected.startswith("yo"):
-                input_lang = "yoruba"
-            elif detected.startswith("ha"):
-                input_lang = "hausa"
-            elif detected.startswith("ig"):
-                input_lang = "igbo"
-            elif detected.startswith("en"):
-                input_lang = "english"
-            # else leave user selection
-        except Exception as e:
-            print("⚠️ Language detection failed:", e)
-
-    elif text_input:
-        input_text = text_input.strip()
-
-    return translate_and_speak(input_text, input_lang, output_lang, use_coqui, clone_voice)
-
-
-# --- Gradio App ---
-with gr.Blocks() as demo:
-    gr.Markdown("## 🌍 Nigerian Voice Translator Assistant")
+with gr.Blocks(title="🌍 Two-Way Translation Assistant") as demo:
+    gr.Markdown("# 🌍 Nigerian Two-Way Voice Translator")
+    gr.Markdown("Speak in your language, hear it in theirs. Supports English ↔ Nigerian languages.")
 
     with gr.Row():
-        input_lang = gr.Dropdown(
-            choices=["auto", "yoruba", "hausa", "igbo", "pidgin", "esan", "tiv", "calabar", "benin", "english"],
-            value="auto",
-            label="Input Language (auto-detect by default)"
-        )
-        output_lang = gr.Dropdown(
-            choices=["english","yoruba","hausa","igbo","pidgin","esan","tiv","calabar","benin"],
-            value="english",
-            label="Output Language"
-        )
-
-    with gr.Tab("🎙️ Voice Input"):
-        mic_input = gr.Audio(sources=["microphone"], type="filepath", label="Speak here")
-
-    with gr.Tab("⌨️ Text Input"):
-        text_input = gr.Textbox(label="Enter text")
-
-    use_coqui = gr.Checkbox(label="Use Coqui TTS (natural accents)", value=True)
-    clone_voice = gr.Checkbox(label="Clone my voice if available", value=False)
-
-    translate_btn = gr.Button("Translate & Speak")
-
-    output_text = gr.Textbox(label="Translation")
-    output_audio = gr.Audio(label="Spoken Output", type="filepath")
+        with gr.Column():
+            gr.Markdown("### 🧑 Speaker A")
+            src_lang = gr.Dropdown(LANGUAGES, value="english", label="Speaker A Language")
+            audio_in_a = gr.Audio(sources=["microphone"], type="filepath", label="🎤 Speak here")
+            translated_a = gr.Textbox(label="Translated Text", interactive=False)
+            audio_out_a = gr.Audio(label="🔊 Translation Audio")
+
+        with gr.Column():
+            gr.Markdown("### 👩 Speaker B")
+            tgt_lang = gr.Dropdown(LANGUAGES, value="yoruba", label="Speaker B Language")
+            audio_in_b = gr.Audio(sources=["microphone"], type="filepath", label="🎤 Reply here")
+            translated_b = gr.Textbox(label="Translated Text", interactive=False)
+            audio_out_b = gr.Audio(label="🔊 Translation Audio")
+
+    clone_voice = gr.Checkbox(value=False, label="🎙️ Use my cloned voice (if my_voice.wav exists)")
+
+    # Wire up A -> B
+    audio_in_a.change(
+        handle_conversation,
+        inputs=[audio_in_a, src_lang, tgt_lang, clone_voice],
+        outputs=[translated_a, audio_out_a]
+    )
 
-    translate_btn.click(
-        fn=handle_input,
-        inputs=[mic_input, text_input, input_lang, output_lang, use_coqui, clone_voice],
-        outputs=[output_text, output_audio]
+    # Wire up B -> A
+    audio_in_b.change(
+        handle_conversation,
+        inputs=[audio_in_b, tgt_lang, src_lang, clone_voice],
+        outputs=[translated_b, audio_out_b]
     )
 
-if __name__ == "__main__":
-    demo.launch()
+demo.launch()