Update app.py

app.py

@@ -255,18 +255,64 @@ def removeCategory(index):
     del st.session_state.categories[index]
     for fname in st.session_state.categorySelect:
         del st.session_state.categorySelect[fname][index]
+
+def _global_rename_key(index):
+    return f"grename_speakers_{index}"
+
+def applyGlobalRenames():
+    """Write all globalRenames entries into speakerRenames and refresh widget keys."""
+    # Clear all existing renames first, then re-apply so removals take effect
+    for fname in st.session_state.speakerRenames:
+        st.session_state.speakerRenames[fname] = {}
+    for entry in st.session_state.globalRenames:
+        display_name = entry["name"]
+        for token in entry["speakers"]:
+            # token format: "filename: SPEAKER_##"
+            if ": " not in token:
+                continue
+            fname, raw_sp = token.split(": ", 1)
+            if fname in st.session_state.speakerRenames:
+                st.session_state.speakerRenames[fname][raw_sp] = display_name
+    # Refresh rename widget keys for the currently viewed file
+    curr = st.session_state.get("select_currFile")
+    if curr and curr in st.session_state.speakerRenames:
+        saved = st.session_state.speakerRenames[curr]
+        results = st.session_state.results.get(curr)
+        if results:
+            for sp in results[0].labels():
+                wk = f"rename_{curr}_{sp}"
+                st.session_state[wk] = saved.get(sp, "")
+
+def addGlobalRename():
+    new_name = st.session_state.globalRenameInput.strip()
+    if not new_name:
+        return
+    st.toast(f"Adding rename '{new_name}'")
+    st.session_state.globalRenames.append({"name": new_name, "speakers": []})
+    st.session_state[_global_rename_key(len(st.session_state.globalRenames) - 1)] = []
+    st.session_state.globalRenameInput = ""
+
+def removeGlobalRename(index):
+    entry = st.session_state.globalRenames[index]
+    st.toast(f"Removing rename '{entry['name']}'")
+    del st.session_state.globalRenames[index]
+    # Rebuild widget keys for remaining entries to stay in sync
+    for i in range(index, len(st.session_state.globalRenames)):
+        next_key = _global_rename_key(i)
+        st.session_state[next_key] = [s for s in st.session_state.globalRenames[i]["speakers"]]
+    applyGlobalRenames()
     
 def updateCategoryOptions(fileName):
     if st.session_state.resetResult:
         return
     currAnnotation, _ = st.session_state.results[fileName]
     speakerNames = list(currAnnotation.labels())
-    # Build reverse map from live rename widget keys: display name -> SPEAKER_##
+    # Build reverse map from speakerRenames (source of truth): display name -> SPEAKER_##
+    saved_renames = st.session_state.speakerRenames.get(fileName, {})
     display_to_raw = {}
     for sp in speakerNames:
-        wk = f"rename_{fileName}_{sp}"
-        live = st.session_state.get(wk, "").strip()
-        display_to_raw[live if live else sp] = sp
+        display = saved_renames.get(sp, sp)
+        display_to_raw[display] = sp
     unusedSpeakers = copy.deepcopy(speakerNames)
     for i, category in enumerate(st.session_state['categories']):
         display_choices = list(st.session_state[f'multiselect_{category}'])
@@ -511,6 +557,8 @@ if 'speakerSegments' not in st.session_state:
     st.session_state.speakerSegments = {}  # {filename: {speaker: [(start,end), ...]}}
 if 'speakerWaveforms' not in st.session_state:
     st.session_state.speakerWaveforms = {} # {filename: (waveform_tensor, sample_rate)}
+if 'globalRenames' not in st.session_state:
+    st.session_state.globalRenames = []    # [{"name": str, "speakers": ["file:SPEAKER_##", ...]}]
 if 'analyzeAllToggle' not in st.session_state:
     st.session_state.analyzeAllToggle = False
     
@@ -753,12 +801,9 @@ try:
 
         unusedSpeakers = st.session_state.unusedSpeakers[currFile]
         categorySelections = st.session_state["categorySelect"][currFile]
-        # Build live raw->display map from rename widget keys (always current)
-        raw_to_display = {}
-        for sp in speakerNames:
-            wk = f"rename_{currFile}_{sp}"
-            live = st.session_state.get(wk, "").strip()
-            raw_to_display[sp] = live if live else sp
+        # Build raw->display map from speakerRenames (source of truth, written by applyGlobalRenames)
+        _saved_renames = st.session_state.speakerRenames.get(currFile, {})
+        raw_to_display = {sp: (_saved_renames.get(sp, sp)) for sp in speakerNames}
         all_speakers_display = [raw_to_display[sp] for sp in speakerNames]
         for i,category in enumerate(st.session_state.categories):
             ms_key = f"multiselect_{category}"
@@ -780,50 +825,79 @@ try:
 
         st.sidebar.divider()
         st.sidebar.subheader("Rename Speakers")
-        st.sidebar.caption("Replace SPEAKER_## labels with real names.")
+        st.sidebar.caption(
+            "Assign a name and select which speaker labels (across all files) it applies to. "
+            "Changes apply to all matched speakers instantly."
+        )
 
-        # --- Speaker clip preview ---
+        # --- Speaker clip preview (identification aid) ---
         file_clips = st.session_state.speakerClips.get(currFile, {})
         if file_clips:
-            st.sidebar.caption(
-                "Listen to a short clip (3–5 s) to help identify each speaker. "
-                "If a clip sounds silent or unclear, press 🔀 to try a different one."
-            )
+            st.sidebar.caption("🎧 Listen to clips to help identify speakers:")
+            current_renames = st.session_state.speakerRenames[currFile]
+            for sp in speakerNames:
+                widget_key = f"rename_{currFile}_{sp}"
+                if widget_key not in st.session_state:
+                    st.session_state[widget_key] = current_renames.get(sp, "")
+                live_name = st.session_state[widget_key].strip()
+                display_label = live_name if live_name else sp
+                st.sidebar.markdown(f"**{display_label}**")
+                if sp in file_clips:
+                    st.sidebar.audio(file_clips[sp], format="audio/wav")
+                    sp_segs = st.session_state.speakerSegments.get(currFile, {}).get(sp, [])
+                    has_waveform = currFile in st.session_state.speakerWaveforms
+                    if has_waveform and len(sp_segs) >= 1:
+                        if st.sidebar.button(
+                            "🔀 Try Another Clip",
+                            key=f"randomize_{currFile}_{sp}",
+                            help="Pick a random clip from a different part of this speaker's audio",
+                        ):
+                            randomize_speaker_clip(currFile, sp)
+                            st.rerun()
+
+        # Build the full list of "filename: SPEAKER_##" tokens across all analyzed files
+        all_speaker_tokens = []
+        for fn in st.session_state.file_names:
+            if fn in st.session_state.results and len(st.session_state.results[fn]) == 2:
+                ann, _ = st.session_state.results[fn]
+                for sp in ann.labels():
+                    all_speaker_tokens.append(f"{fn}: {sp}")
 
-        current_renames = st.session_state.speakerRenames[currFile]
-        for sp in speakerNames:
-            widget_key = f"rename_{currFile}_{sp}"
-            # Seed the widget state on first render only; updateMultiSelect handles
-            # restoring it unconditionally whenever the user switches files.
-            if widget_key not in st.session_state:
-                st.session_state[widget_key] = current_renames.get(sp, "")
-            live_name = st.session_state[widget_key].strip()
-            display_label = live_name if live_name else sp
-            st.sidebar.markdown(f"**{display_label}**")
-            if sp in file_clips:
-                st.sidebar.audio(file_clips[sp], format="audio/wav")
-                # Only show randomize button if there are multiple segments to draw from
-                sp_segs = st.session_state.speakerSegments.get(currFile, {}).get(sp, [])
-                has_waveform = currFile in st.session_state.speakerWaveforms
-                if has_waveform and len(sp_segs) >= 1:
-                    if st.sidebar.button(
-                        "🔀 Try Another Clip",
-                        key=f"randomize_{currFile}_{sp}",
-                        help="Pick a random clip from a different part of this speaker's audio",
-                    ):
-                        randomize_speaker_clip(currFile, sp)
-                        st.rerun()
-            # Label is always the fixed original sp so Streamlit never recreates the widget
-            new_name = st.sidebar.text_input(
-                sp,
-                placeholder="e.g. John",
-                key=widget_key,
-                label_visibility="collapsed"
+        st.sidebar.divider()
+
+        # --- Render existing global rename entries ---
+        def _on_grename_change(idx):
+            key = _global_rename_key(idx)
+            st.session_state.globalRenames[idx]["speakers"] = list(st.session_state[key])
+            applyGlobalRenames()
+
+        for idx, entry in enumerate(st.session_state.globalRenames):
+            grkey = _global_rename_key(idx)
+            if grkey not in st.session_state:
+                st.session_state[grkey] = list(entry["speakers"])
+            st.sidebar.markdown(f"**{entry['name']}**")
+            st.sidebar.multiselect(
+                f"Speakers for {entry['name']}",
+                options=all_speaker_tokens,
+                key=grkey,
+                on_change=_on_grename_change,
+                args=(idx,),
+                label_visibility="collapsed",
             )
-            if new_name.strip():
-                st.session_state.speakerRenames[currFile][sp] = new_name.strip()
-            elif sp in st.session_state.speakerRenames[currFile]:
-                del st.session_state.speakerRenames[currFile][sp]
+            st.sidebar.button(
+                f"Remove '{entry['name']}'",
+                key=f"remove_grename_{idx}",
+                on_click=removeGlobalRename,
+                args=(idx,),
+            )
+
+        # --- Add new global rename ---
+        st.sidebar.text_input(
+            "Add rename",
+            placeholder="e.g. John",
+            key="globalRenameInput",
+            on_change=addGlobalRename,
+        )
 
         catTypeColors = su.colorsCSS(3)
         allColors = su.colorsCSS(len(speakerNames)+len(st.session_state.categories))