Update app.py

app.py

@@ -5,12 +5,9 @@ import torch
 import librosa
 import time
 import os
-import uuid
 import tempfile
 from pydub import AudioSegment
 from transformers import pipeline, AutoFeatureExtractor, AutoModelForAudioClassification
-from scipy.io.wavfile import write
-import sounddevice as sd
 
 # إعداد نموذج Whisper للتفريغ
 whisper_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-base")
@@ -25,14 +22,6 @@ label_map = {
     14: "US"
 }
 
-# تسجيل الصوت من المايكروفون
-def record_audio(duration=5, fs=16000):
-    recording = sd.rec(int(duration * fs), samplerate=fs, channels=1, dtype='int16')
-    sd.wait()
-    temp_path = tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name
-    write(temp_path, fs, recording)
-    return temp_path
-
 # تحويل الصوت إلى النص باستخدام Whisper
 def transcribe_audio(audio_path):
     result = whisper_pipeline(audio_path)
@@ -47,10 +36,9 @@ def predict_accent(audio_path):
         predicted_id = torch.argmax(logits, dim=-1).item()
     return label_map.get(predicted_id, f"Unknown (ID: {predicted_id})")
 
-# معالجة كاملة: تسجيل ← توقع اللهجة ← التفريغ
-def agent_run(duration):
+# المعالجة الكاملة عند إدخال ملف صوتي
+def agent_run(audio_path):
     start_time = time.time()
-    audio_path = record_audio(duration)
     accent = predict_accent(audio_path)
     transcription = transcribe_audio(audio_path)
     elapsed = round(time.time() - start_time, 2)
@@ -59,19 +47,19 @@ def agent_run(duration):
 # واجهة Gradio
 with gr.Blocks() as demo:
     gr.Markdown("## 🧠 Accent Detection Agent")
-    gr.Markdown("Record your voice and let the AI detect the accent and transcribe the speech.")
+    gr.Markdown("Record or upload audio and let the AI detect the accent and transcribe your speech.")
 
-    duration = gr.Slider(3, 10, value=5, label="Recording Duration (seconds)")
-    run_button = gr.Button("🎙️ Start Agent")
+    audio_input = gr.Audio(type="filepath", label="🎙️ Upload or Record Audio", source="microphone")
+    run_button = gr.Button("Analyze")
 
-    audio_output = gr.Audio(label="Recorded Audio")
+    audio_output = gr.Audio(label="Audio")
     accent_output = gr.Textbox(label="Detected Accent")
     transcription_output = gr.Textbox(label="Transcription")
     time_output = gr.Textbox(label="Processing Time")
 
     run_button.click(
         fn=agent_run,
-        inputs=[duration],
+        inputs=[audio_input],
         outputs=[audio_output, accent_output, transcription_output, time_output]
     )