fix(speakers): clarify parse progress and shorten nameplates

- Mark harvested files as Scanned until vocab write finishes
- Prompt and normalize speakers into short title-case nameplates
- Keep speaker finalize chunking on the Settings batch size

data/translation_contexts.json

@@ -3,6 +3,7 @@
   "names": {
     "npc": "Reply with the {language} translation of the NPC name.",
     "npc_only": "Reply with only the {language} translation of the NPC name",
+    "speaker": "Reply with only a short {language} dialogue nameplate for this speaker (e.g. Clerk, Townsman, Electrician). Prefer 1-3 words with title capitalization. Never write a sentence, role description, or explanation.",
     "enemy": "Reply with only the {language} translation of the enemy NPC name",
     "location": "Reply with only the {language} translation of the RPG location name",
     "location_short": "Reply with the {language} translation of the location name.",

gui/translation_tab.py

@@ -199,6 +199,7 @@ class TranslationWorker(QThread):
     progress_signal = pyqtSignal(int, int, str)  # current_file, total_files, filename
     item_progress_signal = pyqtSignal(str, int, int)  # filename, current_item, total_items (for tqdm within file)
     file_error_signal = pyqtSignal(str, str)  # filename, error_message
+    status_signal = pyqtSignal(str)  # updates the top translating_label from the worker
     finished_signal = pyqtSignal(bool, str)
     batch_phase_signal = pyqtSignal(str, object)  # phase name, optional payload
     
@@ -519,7 +520,7 @@ class TranslationWorker(QThread):
             try:
                 from modules.rpgmakermvmz import (
                     handleMVMZ as handler, setSpeakerParseMode, finalizeSpeakerParse,
-                    resetSpeakerState, TOKENS, calculateCost, MODEL,
+                    resetSpeakerState, collectedSpeakerCount, TOKENS, calculateCost, MODEL,
                 )
             except Exception as e:
                 self.emit_log(f"❌ Could not import rpgmakermvmz for speaker-parse: {e}")
@@ -550,12 +551,32 @@ class TranslationWorker(QThread):
                     completed_count += 1
                     self.emit_progress(completed_count, total_files, filename)
 
+            try:
+                speaker_n = collectedSpeakerCount()
+            except Exception:
+                speaker_n = 0
+            if self.should_stop:
+                self.emit_log("⏹ Parse Speakers stopped before translating collected names.")
+            elif speaker_n:
+                self.status_signal.emit(f"Translating {speaker_n} speakers…")
+                self.emit_log(
+                    f"🔤 File scan complete ({completed_count}/{total_files}). "
+                    f"Translating {speaker_n} unique speakers into vocab.txt…"
+                )
+            else:
+                self.status_signal.emit("No speakers found")
+                self.emit_log(
+                    f"🔤 File scan complete ({completed_count}/{total_files}). "
+                    "No speaker names collected."
+                )
+
             try:
                 before_in, before_out = (0, 0)
                 try:
                     before_in, before_out = (int(TOKENS[0]), int(TOKENS[1]))
                 except Exception:
                     pass
+                if not self.should_stop:
                     finalizeSpeakerParse()
                 after_in, after_out = (0, 0)
                 try:
@@ -2486,7 +2507,7 @@ class TranslationTab(QWidget):
             counts[st] = counts.get(st, 0) + 1
         parts = [f"{total} file(s)"]
         for key in (
-            "Done", "Collected", "Writing", "Scanning", "Translating",
+            "Done", "Scanned", "Collected", "Writing", "Scanning", "Translating",
             "Failed", "Skipped", "Unsupported", "Waiting",
         ):
             if counts.get(key):
@@ -2581,10 +2602,15 @@ class TranslationTab(QWidget):
         except Exception:
             pass
         phase = getattr(self, "_batch_ui_phase", None)
+        parse_speakers = bool(
+            getattr(getattr(self, "translation_worker", None), "parse_speakers", False)
+        )
         if getattr(self, "_batch_active", False) and phase == "collect":
             status, color = "Scanning", "#007acc"
         elif getattr(self, "_batch_active", False) and phase == "consume":
             status, color = "Writing", "#007acc"
+        elif parse_speakers:
+            status, color = "Scanning", "#007acc"
         else:
             status, color = "Translating", "#007acc"
         self._set_progress_row(
@@ -2967,6 +2993,7 @@ class TranslationTab(QWidget):
             self.translation_worker.log_signal.connect(self.append_log)
             self.translation_worker.progress_signal.connect(self.update_file_progress)
             self.translation_worker.item_progress_signal.connect(self.update_item_progress)
+            self.translation_worker.status_signal.connect(self.translating_label.setText)
             self.translation_worker.file_error_signal.connect(self.on_file_error)
             self.translation_worker.finished_signal.connect(self.on_translation_finished)
             self.translation_worker.batch_phase_signal.connect(self._on_batch_phase)
@@ -3192,6 +3219,9 @@ class TranslationTab(QWidget):
         """Update the file-level progress."""
         batch_active = getattr(self, "_batch_active", False)
         phase = getattr(self, "_batch_ui_phase", None) if batch_active else None
+        parse_speakers = bool(
+            getattr(getattr(self, "translation_worker", None), "parse_speakers", False)
+        )
 
         if batch_active:
             if phase == "collect":
@@ -3208,11 +3238,29 @@ class TranslationTab(QWidget):
                 return
 
         self.files_completed = current_file
+        if parse_speakers:
+            self.files_translated_label.setText(f"{current_file}/{total_files} scanned")
+        else:
             self.files_translated_label.setText(f"{current_file}/{total_files}")
 
         # Row details (tokens/cost) come from parsed stdout via append_log. If that
         # did not run (older modules / parse miss), finalize so the row is not stuck.
         if filename in self.file_progress_items:
+            if parse_speakers:
+                # Harvest only - translation of nameplates happens after all files.
+                self.mark_file_queued(filename)
+                meta = self.file_progress_items[filename]
+                try:
+                    meta["label"].setText("Scanned")
+                except Exception:
+                    pass
+                self._set_progress_row(
+                    filename,
+                    status="Scanned",
+                    status_color="#d4a017",
+                    progress="names",
+                )
+            else:
                 sl = self.file_progress_items[filename].get("status_label")
                 if not sl or not sl.text():
                     self.mark_file_complete(filename, success=True)
@@ -3232,10 +3280,15 @@ class TranslationTab(QWidget):
             )
             self.batch_live_status.setText(f"Current file: {filename}")
         if current_file < total_files:
-            self.translating_label.setText("Translating...")
+            self.translating_label.setText(
+                "Scanning speakers..." if parse_speakers else "Translating..."
+            )
         else:
             if batch_active and phase == "consume":
                 self.translating_label.setText("Finishing batch…")
+            elif parse_speakers:
+                # Keep a clear post-scan state until status_signal / finished.
+                self.translating_label.setText("Translating speakers…")
             else:
                 self.translating_label.setText("—")
     
@@ -3322,6 +3375,15 @@ class TranslationTab(QWidget):
         """Apply UI changes for a finished translation run."""
         if getattr(self, "_batch_active", False):
             self._on_batch_phase("done", None)
+        # Parse Speakers: promote Scanned rows to Done only after vocab write finishes.
+        try:
+            worker = getattr(self, "translation_worker", None)
+            if success and worker and getattr(worker, "parse_speakers", False):
+                for fname, meta in (getattr(self, "file_progress_items", {}) or {}).items():
+                    if (meta.get("status_text") or "") == "Scanned":
+                        self.mark_file_complete(fname, success=True)
+        except Exception:
+            pass
         # Hide the stop button and show the reset/back button
         try:
             self.stop_button.setVisible(False)

modules/rpgmakermvmz.py

@@ -4762,6 +4762,35 @@ def _is_plausible_speaker(name: str) -> bool:
     return True
 
 
+def _normalize_speaker_nameplate(translated: str) -> str:
+    """Normalize a model TL into a short dialogue nameplate."""
+    out = (translated or "").strip()
+    out = re.sub(r"^(speaker:\s*)", "", out, flags=re.IGNORECASE)
+    out = out.strip("'\"“”‘’「」")
+    # If the model ignored instructions and wrote a sentence, keep the lead clause.
+    if re.search(r"[.!?]", out):
+        out = re.split(r"[.!?]", out, maxsplit=1)[0].strip()
+    # Drop trailing commas / "the kind you'd..." style appositives.
+    if "," in out:
+        out = out.split(",", 1)[0].strip()
+    out = re.sub(r"\s+", " ", out).strip()
+    words = out.split()
+    # Descriptive expansions often start with filler; keep the trailing role words.
+    if len(words) > 3 and re.match(
+        r"^(JUST|ONLY|THE|A|AN|YOUR|SOME|ANY|THIS|THAT)\b", words[0], re.I
+    ):
+        words = words[-2:]
+    elif len(words) > 4:
+        words = words[:3]
+    out = " ".join(words).title().replace("'S", "'s")
+    out = re.sub(
+        r"(\d)(St|Nd|Rd|Th)\b",
+        lambda m: m.group(1) + m.group(2).lower(),
+        out,
+    )
+    return out
+
+
 # Save some money and enter the character before translation
 def getSpeaker(speaker: str):
     """Return (and possibly collect) speaker name.
@@ -4799,15 +4828,14 @@ def getSpeaker(speaker: str):
         pass
     response = translateAI(
         speaker,
-        ctx("names.npc"),
+        ctx("names.speaker"),
         False,
     )
     try:
         THREAD_CTX.in_speaker = False
     except Exception:
         pass
-    translated = response[0].strip().title().replace("'S", "'s").replace("Speaker: ", "")
+    translated = _normalize_speaker_nameplate(response[0])
-    translated = re.sub(r'(\d)(St|Nd|Rd|Th)\b', lambda m: m.group(1) + m.group(2).lower(), translated)
 
     if re.search(r"([a-zA-Z？?])", translated) is None:
         try:
@@ -4816,15 +4844,14 @@ def getSpeaker(speaker: str):
             pass
         response = translateAI(
             speaker,
-            ctx("names.npc"),
+            ctx("names.speaker"),
             False,
         )
         try:
             THREAD_CTX.in_speaker = False
         except Exception:
             pass
-        translated = response[0].strip().title().replace("'S", "'s")
+        translated = _normalize_speaker_nameplate(response[0])
-        translated = re.sub(r'(\d)(St|Nd|Rd|Th)\b', lambda m: m.group(1) + m.group(2).lower(), translated)
 
     with _speakerCacheLock:
         if speaker not in _speakerCache:
@@ -4889,6 +4916,11 @@ def translateAI(text, history, history_ctx=None):
     
     # Update config estimate mode based on global ESTIMATE
     TRANSLATION_CONFIG.estimateMode = bool(ESTIMATE)
+    # Keep batch size in sync with Settings / .env (module import can be stale).
+    try:
+        TRANSLATION_CONFIG.batchSize = getPricingConfig(MODEL)["batchSize"]
+    except Exception:
+        pass
     
     # Call the new shared translation function
     # Prefer thread-local filename for logging; fall back to global
@@ -5009,6 +5041,12 @@ def setSpeakerParseMode(flag: bool):
     global SPEAKER_PARSE_MODE
     SPEAKER_PARSE_MODE = bool(flag)
 
+def collectedSpeakerCount() -> int:
+    """Unique speaker names waiting for (or already in) the parse-mode glossary."""
+    with _speakerCacheLock:
+        return len(SPEAKER_COLLECTED)
+
+
 def finalizeSpeakerParse():
     """Batch translate collected speakers and write fresh # Speakers section."""
     if not SPEAKER_PARSE_MODE:
@@ -5025,9 +5063,10 @@ def finalizeSpeakerParse():
                 THREAD_CTX.in_speaker = True
             except Exception:
                 pass
+            # Full list; translateAI chunks with the Settings batch size.
             resp = translateAI(
                 to_translate,
-                ctx("names.npc"),
+                ctx("names.speaker"),
                 True,
             )
             try:
@@ -5057,9 +5096,9 @@ def finalizeSpeakerParse():
             tl_list = resp[0]
             with _speakerCacheLock:
                 for orig, tl in zip(to_translate, tl_list):
-                    norm = tl.title().replace("'S", "'s").replace("Speaker: ", "")
+                    norm = _normalize_speaker_nameplate(tl)
                     if re.search(r"([a-zA-Z？?])", norm) is None:
-                        norm = tl  # keep raw if heuristic fails
+                        norm = (tl or "").strip() or orig
                     if orig not in _speakerCache:
                         _speakerCache[orig] = norm
                         NAMESLIST.append([orig, norm])

tests/test_speaker_nameplate.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Tests for RPG Maker speaker nameplate normalization."""
+
+import unittest
+
+from modules.rpgmakermvmz import _normalize_speaker_nameplate
+
+
+class SpeakerNameplateNormalizeTests(unittest.TestCase):
+    def test_title_case_short_name(self):
+        self.assertEqual(_normalize_speaker_nameplate("clerk"), "Clerk")
+
+    def test_strips_speaker_prefix(self):
+        self.assertEqual(_normalize_speaker_nameplate("Speaker: Townsman"), "Townsman")
+
+    def test_collapses_descriptive_sentence(self):
+        raw = "Just your average electrician guy, the kind you'd find anywhere."
+        self.assertEqual(_normalize_speaker_nameplate(raw), "Electrician Guy")
+
+    def test_truncates_long_phrase(self):
+        self.assertEqual(
+            _normalize_speaker_nameplate("ELECTRICIAN GUY FROM NEXT DOOR SHOP"),
+            "Electrician Guy From",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()