Update app.py

app.py

@@ -151,12 +151,8 @@ def generate_speaker_clips(annotations, waveform, sample_rate, file_index):
 
         clip_waveform = waveform[:, start_sample:end_sample]
 
-        # Use soundfile (pure C, no FFmpeg/torchcodec) to encode WAV bytes
-        # soundfile expects (samples, channels) numpy array; waveform is (channels, samples) tensor
-        clip_np = clip_waveform.numpy().T
         buffer = io.BytesIO()
-        import soundfile as sf
-        sf.write(buffer, clip_np, sample_rate, format="WAV", subtype="PCM_16")
+        torchaudio.save(buffer, clip_waveform, sample_rate, format="wav")
         buffer.seek(0)
         clips[speaker] = buffer.read()
 
@@ -680,16 +676,19 @@ try:
 
         current_renames = st.session_state.speakerRenames[currFileIndex]
         for sp in speakerNames:
-            display_label = current_renames.get(sp, sp)
+            # Read the live widget value so label updates immediately on rename
+            widget_key = f"rename_{currFileIndex}_{sp}"
+            live_name = st.session_state.get(widget_key, "").strip()
+            display_label = live_name if live_name else current_renames.get(sp, sp)
             st.sidebar.markdown(f"**{display_label}**")
             if sp in file_clips:
                 st.sidebar.audio(file_clips[sp], format="audio/wav")
             current_label = current_renames.get(sp, "")
             new_name = st.sidebar.text_input(
-                f"Rename {sp}",
+                f"Rename {display_label}",
                 value=current_label,
-                placeholder=f"e.g. John",
-                key=f"rename_{currFileIndex}_{sp}"
+                placeholder="e.g. John",
+                key=widget_key
             )
             if new_name.strip():
                 st.session_state.speakerRenames[currFileIndex][sp] = new_name.strip()