diff --git a/.gitignore b/.gitignore
index 2d1a2e4..9e3a0e9 100644
--- a/.gitignore
+++ b/.gitignore
@@ -34,7 +34,8 @@ json/*
 !json/.gitkeep
 docs/*
 !docs/PARROT.md
-Audio/
+Audio/*
+!Audio/.gitkeep
 
 # -----------------------------------------------------------------------------
 # Python
diff --git a/Audio/.gitkeep b/Audio/.gitkeep
new file mode 100644
index 0000000..e69de29
diff --git a/README.md b/README.md
index 6554782..9c51d11 100644
--- a/README.md
+++ b/README.md
@@ -23,7 +23,9 @@ Copy `adn-server.example.yaml` to `adn-server.yaml` and edit with your settings.
 
 Voice features (announcements, TTS, recording) use a separate config file. Copy `adn-voice.example.yaml` to `adn-voice.yaml` and edit. If the file does not exist, voice features are disabled (no error). Changes are hot-reloaded every 15 seconds.
 
-TTS (Text-to-Speech) requires **ffmpeg** installed on the system (see Requirements above). The pipeline is: `.txt` → gTTS → `.mp3` → ffmpeg → `.wav` → vocoder → `.ambe`
+- **Each item** (ANNOUNCEMENTS, TTS_ANNOUNCEMENTS) has its own `LANGUAGE` and `ENABLED: true` to activate.
+- **ANNOUNCEMENT_LANGUAGES** is optional (for voice ident only); announcements/TTS work without it.
+- **TTS** requires ffmpeg + vocoder (TTS_VOCODER_CMD or TTS_AMBESERVER_HOST). Pipeline: `.txt` → gTTS → `.mp3` → ffmpeg → `.wav` → vocoder → `.ambe`
 
 ## Run
 
diff --git a/adn-voice.example.yaml b/adn-voice.example.yaml
index 1b625f0..a38b988 100644
--- a/adn-voice.example.yaml
+++ b/adn-voice.example.yaml
@@ -3,10 +3,12 @@
 #
 # If adn-voice.yaml does not exist, voice features are disabled (no error).
 # This file is re-checked every 15 s; changes take effect without restart.
+#
+# Each announcement/TTS has its own LANGUAGE. ENABLED: true activates that item.
 
 VOICE:
-  # Languages available for voice announcements (comma-separated, e.g. "es_ES,en_GB").
-  # Leave empty to disable voice features.
+  # Optional: languages for voice ident (MASTER systems with VOICE_IDENT).
+  # If empty, ident uses per-system ANNOUNCEMENT_LANGUAGE. Not needed for announcements/TTS.
   ANNOUNCEMENT_LANGUAGES: ""
 
   # Base path for AMBE audio files (relative to project root).
@@ -17,22 +19,23 @@ VOICE:
   # Pre-recorded .ambe files played on a talkgroup at intervals or hourly.
   # FILE: filename without .ambe, looked up in Audio/<LANGUAGE>/ondemand/
   # MODE: "hourly" (once per hour) or "interval" (every INTERVAL seconds)
+  # ENABLED: true to activate; false to keep configured but inactive
   # ---------------------------------------------------------------------------
   ANNOUNCEMENTS: []
   # Example (all variables used):
   # ANNOUNCEMENTS:
-  #   - ENABLED: false
+  #   - ENABLED: true
   #     FILE: announcement
   #     TG: 2
   #     MODE: hourly
   #     INTERVAL: 3600
   #     LANGUAGE: en_GB
-  #   - ENABLED: false
+  #   - ENABLED: true
   #     FILE: announcement2
   #     TG: 2
   #     MODE: interval
-  #     INTERVAL: 1800
-  #     LANGUAGE: en_GB
+  #     INTERVAL: 60
+  #     LANGUAGE: es_ES
 
   # ---------------------------------------------------------------------------
   # Recording
@@ -50,6 +53,7 @@ VOICE:
   # TTS (Text-to-Speech) — array
   # Converts .txt files to AMBE via: gTTS -> ffmpeg -> vocoder/AMBEServer
   # FILE: filename without .txt, looked up in Audio/<LANGUAGE>/ondemand/
+  # ENABLED: true to activate; requires TTS_VOCODER_CMD or TTS_AMBESERVER_HOST
   # ---------------------------------------------------------------------------
 
   # External vocoder command ({wav} and {ambe} are replaced at runtime).
@@ -70,15 +74,15 @@ VOICE:
 
   # Example:
   # TTS_ANNOUNCEMENTS:
-  #   - ENABLED: false
+  #   - ENABLED: true
   #     FILE: text1
   #     TG: 2
   #     MODE: hourly
   #     INTERVAL: 3600
   #     LANGUAGE: en_GB
-  #   - ENABLED: false
+  #   - ENABLED: true
   #     FILE: text2
   #     TG: 2
   #     MODE: interval
-  #     INTERVAL: 3600
-  #     LANGUAGE: en_GB
+  #     INTERVAL: 60
+  #     LANGUAGE: es_ES
diff --git a/src/adn_server/application/ident_use_cases.py b/src/adn_server/application/ident_use_cases.py
index 68118f2..645439b 100644
--- a/src/adn_server/application/ident_use_cases.py
+++ b/src/adn_server/application/ident_use_cases.py
@@ -65,6 +65,13 @@ class IdentUseCases:
         systems_cfg = self._config.get("SYSTEMS", {})
         protocols = self._get_protocols()
         ann_lang = (self._config.get("VOICE", {}).get("ANNOUNCEMENT_LANGUAGES") or "").strip()
+        if not ann_lang:
+            # Fallback: derive from MASTER systems with VOICE_IDENT (each has ANNOUNCEMENT_LANGUAGE)
+            langs = set()
+            for sys_cfg in systems_cfg.values():
+                if sys_cfg.get("MODE") == "MASTER" and sys_cfg.get("VOICE_IDENT"):
+                    langs.add(sys_cfg.get("ANNOUNCEMENT_LANGUAGE", "en_GB"))
+            ann_lang = ",".join(sorted(langs)) if langs else ""
         if not ann_lang:
             return
         words_by_lang = self._voice.get_ambe_words(ann_lang, self._audio_path)
diff --git a/src/adn_server/application/voice_use_cases.py b/src/adn_server/application/voice_use_cases.py
index 9626eeb..da0329c 100644
--- a/src/adn_server/application/voice_use_cases.py
+++ b/src/adn_server/application/voice_use_cases.py
@@ -534,12 +534,9 @@ class VoiceUseCases:
         protocol = self._get_protocols().get(system)
         if not protocol or not getattr(protocol, "STATUS", None):
             return
-        ann_lang = (self._config.get("VOICE", {}).get("ANNOUNCEMENT_LANGUAGES") or "").strip()
-        if not ann_lang:
-            return
-        words_by_lang = self.get_ambe_words(ann_lang, self._audio_path)
         sys_cfg = self._config.get("SYSTEMS", {}).get(system, {})
         _lang = sys_cfg.get("ANNOUNCEMENT_LANGUAGE", "en_GB")
+        words_by_lang = self.get_ambe_words(_lang, self._audio_path)
         if _lang not in words_by_lang:
             return
         words = words_by_lang[_lang]
diff --git a/src/adn_server/main.py b/src/adn_server/main.py
index df229ac..170b0c4 100644
--- a/src/adn_server/main.py
+++ b/src/adn_server/main.py
@@ -234,10 +234,12 @@ def main() -> None:
 
     recording_handler = RecordingHandler(config, project_root)
 
-    # Voice: AMBE words + pkt_gen (legacy readAMBE + mk_voice). Use Default when languages + audio path set.
+    # Voice: AMBE words + pkt_gen (legacy readAMBE + mk_voice). Use Default when Audio dir exists.
+    # ANNOUNCEMENT_LANGUAGES is only for voice ident; announcements/TTS use per-item LANGUAGE.
     audio_path = os.path.join(project_root, config.get("VOICE", {}).get("AUDIO_PATH", "Audio"))
-    ann_lang = (config.get("VOICE", {}).get("ANNOUNCEMENT_LANGUAGES") or "").strip()
-    if ann_lang and os.path.isdir(audio_path):
+    if config.get("VOICE") and not os.path.isdir(audio_path):
+        os.makedirs(audio_path, exist_ok=True)
+    if os.path.isdir(audio_path):
         voice_provider = DefaultVoiceProvider()
     else:
         voice_provider = StubVoiceProvider()