Update app.py

app.py

@@ -5,28 +5,25 @@ from pydub import AudioSegment
 import numpy as np
 import io
 
-# Lightweight Whisper (CPU-safe)
+# Load Whisper TINY (multilingual) – CPU safe
 asr = pipeline(
     "automatic-speech-recognition",
     model="openai/whisper-tiny",
     device="cpu"
 )
 
-def speech_to_speech_translation(audio):
-    if audio is None:
+def speech_to_speech_translation(audio_path):
+    if audio_path is None:
         return None
 
-    sr, audio_array = audio
-
-    # Speech → French text (NON-ENGLISH)
+    # 1. Speech file → French text
     result = asr(
-        audio_array,
-        sampling_rate=sr,
+        audio_path,
         generate_kwargs={"task": "translate", "language": "french"}
     )
     french_text = result["text"]
 
-    # French text → French speech (free TTS)
+    # 2. French text → French speech (free, lightweight)
     tts = gTTS(text=french_text, lang="fr")
     mp3_fp = io.BytesIO()
     tts.write_to_fp(mp3_fp)
@@ -40,8 +37,8 @@ def speech_to_speech_translation(audio):
 
 demo = gr.Interface(
     fn=speech_to_speech_translation,
-    inputs=gr.Audio(type="numpy"),
-    outputs=gr.Audio(type="numpy"),
+    inputs=gr.Audio(type="filepath", label="Upload speech"),
+    outputs=gr.Audio(type="numpy", label="French speech output"),
     title="Speech-to-Speech Translation (French)",
     allow_flagging="never"
 )