Update app.py

app.py

@@ -1,4 +1,3 @@
-
 import os
 import shutil
 import zipfile
@@ -6,7 +5,7 @@ import torch
 from pathlib import Path
 
 import gradio as gr
-from pydub import AudioSegment  # Correction de l'import erroné de 'pubdub'
+from pydub import AudioSegment
 from transformers import pipeline
 
 # -------------------------------------------------
@@ -53,21 +52,35 @@ def validate_segments(audio_path, table_data, metadata_state, word_timestamps):
     updated_metadata = []
 
     for i, row in enumerate(table_data):
-        text, start_time, end_time, segment_id = row
-        if not text or start_time is None or end_time is None:
-            continue
-        
-        segment_id = segment_id or f"seg_{i+1:02d}"
+        if len(row) < 1:
+            continue  # Ignorer les lignes vides
+
+        text = row[0]  # L'utilisateur n'entre que le texte
+        segment_id = f"seg_{i+1:02d}"
+
+        # Trouver les timestamps correspondant au texte
+        matching_timestamps = [
+            (start, end) for word, (start, end) in word_timestamps if word in text
+        ]
+
+        if matching_timestamps:
+            start_time, end_time = matching_timestamps[0]  # Premier match trouvé
+        else:
+            start_time, end_time = None, None  # Impossible de déterminer
+
+        if start_time is None or end_time is None:
+            continue  # Ignorer si aucun timestamp trouvé
+
         start_ms, end_ms = int(float(start_time) * 1000), int(float(end_time) * 1000)
         if start_ms < 0 or end_ms <= start_ms:
             continue
-        
+
         segment_filename = f"{Path(audio_path).stem}_{segment_id}.wav"
         segment_path = os.path.join(TEMP_DIR, segment_filename)
-        
+
         extract = original_audio[start_ms:end_ms]
         extract.export(segment_path, format="wav")
-        
+
         segment_paths.append(segment_path)
         updated_metadata.append({
             "audio_file": segment_filename,
@@ -112,10 +125,10 @@ with gr.Blocks() as demo:
     audio_input = gr.Audio(type="filepath", label="Fichier audio")
     raw_transcription = gr.Textbox(label="Transcription", interactive=False)
     table = gr.Dataframe(
-        headers=["Texte", "Début (s)", "Fin (s)", "ID"],
-        datatype=["str", "number", "number", "str"],
-        row_count=(1, "dynamic"),  # Gestion dynamique confirmée
-        col_count=4
+        headers=["Texte"],
+        datatype=["str"],
+        row_count=(1, "dynamic"),  # Gestion dynamique
+        col_count=1
     )
     validate_button = gr.Button("Valider")
     audio_players = [gr.Audio(label=f"Extrait {i+1}", interactive=False) for i in range(20)]
@@ -129,4 +142,3 @@ with gr.Blocks() as demo:
     generate_button.click(generate_zip, inputs=metadata_state, outputs=zip_file)
 
 demo.queue().launch()
-