Update app.py

app.py

@@ -54,8 +54,11 @@ def transcribe_audio(audio_path):
         return "Erreur : Aucun timestamp détecté.", None, [], ""
     
     raw_transcription = " ".join([w["text"] for w in words])
+
+    # 🔄 Correction typographique AVANT affichage
     raw_transcription = correct_typography(raw_transcription)
-    
+
+    # 🔄 Ajout des timestamps de fin avec marge de sécurité
     MARGIN = 0.06  # 60ms
     word_timestamps = []
 
@@ -63,9 +66,10 @@ def transcribe_audio(audio_path):
         start_time = w["timestamp"][0]
         end_time = w["timestamp"][1] if w["timestamp"][1] is not None else start_time + 0.5
 
+        # Vérifier qu'on ne dépasse pas le début du mot suivant
         if i < len(words) - 1:
             next_start_time = words[i + 1]["timestamp"][0]
-            end_time = min(end_time + MARGIN, next_start_time - 0.01)
+            end_time = min(end_time + MARGIN, next_start_time - 0.01)  # On laisse 10ms de sécurité
 
         word_timestamps.append((w["text"], start_time, end_time))
 
@@ -115,41 +119,43 @@ def save_segments(table_data):
 # -------------------------------------------------
 def generate_zip(metadata_state, audio_path, zip_name):
     if isinstance(metadata_state, tuple): 
-        metadata_state = metadata_state[0]
+        metadata_state = metadata_state[0]  # Extraire le DataFrame si c'est un tuple
     
     if metadata_state is None or metadata_state.empty:
         print("[LOG ERROR] Aucun segment valide trouvé pour la génération du ZIP.")
         return None
-    
     zip_folder_name = f"{zip_name}_dataset"
     zip_path = os.path.join(TEMP_DIR, f"{zip_folder_name}.zip")
-    metadata_csv_path = os.path.join(TEMP_DIR, "metadata.csv")
-    
-    metadata_state["ID"] = [f"{zip_name}_seg_{i+1:02d}" for i in range(len(metadata_state))]
-    metadata_state["Texte"] = metadata_state["Texte"].apply(correct_typography)
-    metadata_state["Commentaires"] = ""
-    metadata_state.to_csv(metadata_csv_path, sep="|", index=False)
-    
-    with zipfile.ZipFile(zip_path, "w", zipfile.ZIP_DEFLATED) as zf:
-        zf.write(metadata_csv_path, "metadata.csv")
-    
-    print("[LOG] Fichier ZIP généré avec succès.")
-    return zip_path
-    if metadata_state is None or metadata_state.empty:
-        return None
+    if os.path.exists(zip_path):
+        os.remove(zip_path)
     
-    zip_folder_name = f"{zip_name}_dataset"
-    zip_path = os.path.join(TEMP_DIR, f"{zip_folder_name}.zip")
     metadata_csv_path = os.path.join(TEMP_DIR, "metadata.csv")
-    
+
+    # Assurer que les ID et fichiers audio correspondent
     metadata_state["ID"] = [f"{zip_name}_seg_{i+1:02d}" for i in range(len(metadata_state))]
-    metadata_state["Texte"] = metadata_state["Texte"].apply(correct_typography)
+
+    # Ajouter une colonne "Commentaires" vide
     metadata_state["Commentaires"] = ""
+
+    # Réorganiser l’ordre des colonnes
+    metadata_state = metadata_state[["ID", "Texte", "Début (s)", "Fin (s)", "Commentaires"]]
+
+    # Sauvegarde du fichier CSV
     metadata_state.to_csv(metadata_csv_path, sep="|", index=False)
-    
+
     with zipfile.ZipFile(zip_path, "w", zipfile.ZIP_DEFLATED) as zf:
         zf.write(metadata_csv_path, "metadata.csv")
+        original_audio = AudioSegment.from_file(audio_path)
+        
+        for _, row in metadata_state.iterrows():
+            start_ms, end_ms = int(row["Début (s)"] * 1000), int(row["Fin (s)"] * 1000)
+            segment_audio = original_audio[start_ms:end_ms]
+            segment_filename = f"{row['ID']}.wav"
+            segment_path = os.path.join(TEMP_DIR, segment_filename)
+            segment_audio.export(segment_path, format="wav")
+            zf.write(segment_path, segment_filename)
     
+    print("[LOG] Fichier ZIP généré avec succès.")
     return zip_path
 
 # -------------------------------------------------
@@ -158,17 +164,20 @@ def generate_zip(metadata_state, audio_path, zip_name):
 with gr.Blocks() as demo:
     gr.Markdown("# Application de Découpe Audio")
     metadata_state = gr.State(init_metadata_state())
+    
     audio_input = gr.Audio(type="filepath", label="Fichier audio")
     zip_name = gr.Textbox(label="Nom du fichier ZIP", interactive=True)
     raw_transcription = gr.Textbox(label="Transcription", interactive=True)
     transcription_timestamps = gr.Textbox(label="Transcription avec Timestamps", interactive=True)
     table = gr.Dataframe(headers=["Texte", "Début (s)", "Fin (s)"], datatype=["str", "str", "str"], row_count=(1, "dynamic"))
     save_button = gr.Button("Enregistrer les segments")
+    save_message = gr.Markdown(label="📢 **Message de confirmation**")
     generate_button = gr.Button("Générer ZIP")
     zip_file = gr.File(label="Télécharger le ZIP")
+    word_timestamps = gr.State()
     
-    audio_input.change(transcribe_audio, inputs=audio_input, outputs=[raw_transcription, transcription_timestamps, audio_input])
-    save_button.click(save_segments, inputs=table, outputs=[metadata_state])
+    audio_input.change(transcribe_audio, inputs=audio_input, outputs=[raw_transcription, word_timestamps, transcription_timestamps, audio_input])
+    save_button.click(save_segments, inputs=table, outputs=[metadata_state, save_message])
     generate_button.click(generate_zip, inputs=[metadata_state, audio_input, zip_name], outputs=zip_file)
 
-demo.queue().launch()
+demo.queue().launch()