Update app.py

app.py

@@ -1,30 +1,47 @@
+import os
 import gradio as gr
-from stt_engine import STTEngine
+from transformers import pipeline
 from tts_engine import TTSEngine
-from translator import TranslatorEngine
 
-# Init engines
-stt_engine = STTEngine()
+# STT: Whisper
+stt_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-small", device=-1)
+
+# Translation: MarianMT (generic English <-> multilingual)
+translator_pipeline = pipeline("translation", model="Helsinki-NLP/opus-mt-mul-en")
+
 tts_engine = TTSEngine(use_coqui=True)
-translator = TranslatorEngine()
 
 LANGUAGES = [
     "english", "yoruba", "igbo", "hausa", "pidgin",
     "esan", "tiv", "calabar", "benin"
 ]
 
+def transcribe(audio, language="en"):
+    """Speech to text using Whisper."""
+    if audio is None:
+        return ""
+    result = stt_pipeline(audio, generate_kwargs={"language": language})
+    return result["text"]
+
+def translate(text, src_lang, tgt_lang):
+    """Dummy translation with Marian (you can extend with Nigerian mappings)."""
+    if not text:
+        return ""
+    # For Nigerian languages not covered by Marian, just return the same text
+    if src_lang not in ["english", "en"] or tgt_lang not in ["english", "en"]:
+        return f"[{src_lang}->{tgt_lang}] {text}"
+    translated = translator_pipeline(text)[0]["translation_text"]
+    return translated
+
 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)
+    # Step 1: STT
+    text = transcribe(audio, language="en" if src_lang == "english" else None)
 
-    # Translate
-    translated = translator.translate(text, src_lang, tgt_lang)
+    # Step 2: Translate
+    translated = translate(text, src_lang, tgt_lang)
 
-    # TTS (with cloned voice if available)
+    # Step 3: TTS
     audio_path = tts_engine.speak(translated, lang=tgt_lang, voice_clone=clone_voice)
 
     return translated, audio_path