diff --git a/td/telegram/MessageContent.cpp b/td/telegram/MessageContent.cpp
index 5244a9284..5a076c9f5 100644
--- a/td/telegram/MessageContent.cpp
+++ b/td/telegram/MessageContent.cpp
@@ -3722,6 +3722,9 @@ void register_message_content(Td *td, const MessageContent *content, FullMessage
       }
       return;
     }
+    case MessageContentType::VoiceNote:
+      return td->voice_notes_manager_->register_voice_note(static_cast<const MessageVoiceNote *>(content)->file_id,
+                                                           full_message_id, source);
     case MessageContentType::Poll:
       return td->poll_manager_->register_poll(static_cast<const MessagePoll *>(content)->poll_id, full_message_id,
                                               source);
@@ -3750,6 +3753,12 @@ void reregister_message_content(Td *td, const MessageContent *old_content, const
         }
         break;
       }
+      case MessageContentType::VoiceNote:
+        if (static_cast<const MessageVoiceNote *>(old_content)->file_id ==
+            static_cast<const MessageVoiceNote *>(new_content)->file_id) {
+          return;
+        }
+        break;
       case MessageContentType::Poll:
         if (static_cast<const MessagePoll *>(old_content)->poll_id ==
             static_cast<const MessagePoll *>(new_content)->poll_id) {
@@ -3784,6 +3793,9 @@ void unregister_message_content(Td *td, const MessageContent *content, FullMessa
       }
       return;
     }
+    case MessageContentType::VoiceNote:
+      return td->voice_notes_manager_->unregister_voice_note(static_cast<const MessageVoiceNote *>(content)->file_id,
+                                                             full_message_id, source);
     case MessageContentType::Poll:
       return td->poll_manager_->unregister_poll(static_cast<const MessagePoll *>(content)->poll_id, full_message_id,
                                                 source);
diff --git a/td/telegram/VoiceNotesManager.cpp b/td/telegram/VoiceNotesManager.cpp
index 4f069ca01..bdaeb17c8 100644
--- a/td/telegram/VoiceNotesManager.cpp
+++ b/td/telegram/VoiceNotesManager.cpp
@@ -129,12 +129,36 @@ void VoiceNotesManager::create_voice_note(FileId file_id, string mime_type, int3
   on_get_voice_note(std::move(v), replace);
 }
 
-SecretInputMedia VoiceNotesManager::get_secret_input_media(FileId voice_file_id,
+void VoiceNotesManager::register_voice_note(FileId voice_note_file_id, FullMessageId full_message_id,
+                                            const char *source) {
+  if (full_message_id.get_message_id().is_scheduled() || !full_message_id.get_message_id().is_server()) {
+    return;
+  }
+  LOG(INFO) << "Register voice note " << voice_note_file_id << " from " << full_message_id << " from " << source;
+  bool is_inserted = voice_note_messages_[voice_note_file_id].insert(full_message_id).second;
+  LOG_CHECK(is_inserted) << source << ' ' << voice_note_file_id << ' ' << full_message_id;
+  auto voice_note = get_voice_note(voice_note_file_id);
+  CHECK(voice_note != nullptr);
+}
+
+void VoiceNotesManager::unregister_voice_note(FileId voice_note_file_id, FullMessageId full_message_id,
+                                              const char *source) {
+  if (full_message_id.get_message_id().is_scheduled() || !full_message_id.get_message_id().is_server()) {
+    return;
+  }
+  LOG(INFO) << "Unregister voice note " << voice_note_file_id << " from " << full_message_id << " from " << source;
+  auto &message_ids = voice_note_messages_[voice_note_file_id];
+  auto is_deleted = message_ids.erase(full_message_id) > 0;
+  LOG_CHECK(is_deleted) << source << ' ' << voice_note_file_id << ' ' << full_message_id;
+  if (message_ids.empty()) {
+    voice_note_messages_.erase(voice_note_file_id);
+  }
+}
+
+SecretInputMedia VoiceNotesManager::get_secret_input_media(FileId voice_note_file_id,
                                                            tl_object_ptr<telegram_api::InputEncryptedFile> input_file,
                                                            const string &caption, int32 layer) const {
-  auto *voice_note = get_voice_note(voice_file_id);
-  CHECK(voice_note != nullptr);
-  auto file_view = td_->file_manager_->get_file_view(voice_file_id);
+  auto file_view = td_->file_manager_->get_file_view(voice_note_file_id);
   if (!file_view.is_encrypted_secret() || file_view.encryption_key().empty()) {
     return SecretInputMedia{};
   }
@@ -144,6 +168,9 @@ SecretInputMedia VoiceNotesManager::get_secret_input_media(FileId voice_file_id,
   if (!input_file) {
     return SecretInputMedia{};
   }
+
+  auto *voice_note = get_voice_note(voice_note_file_id);
+  CHECK(voice_note != nullptr);
   vector<tl_object_ptr<secret_api::DocumentAttribute>> attributes;
   attributes.push_back(make_tl_object<secret_api::documentAttributeAudio>(
       secret_api::documentAttributeAudio::VOICE_MASK | secret_api::documentAttributeAudio::WAVEFORM_MASK,
diff --git a/td/telegram/VoiceNotesManager.h b/td/telegram/VoiceNotesManager.h
index 0a5fa9668..9f610480b 100644
--- a/td/telegram/VoiceNotesManager.h
+++ b/td/telegram/VoiceNotesManager.h
@@ -7,6 +7,7 @@
 #pragma once
 
 #include "td/telegram/files/FileId.h"
+#include "td/telegram/FullMessageId.h"
 #include "td/telegram/SecretInputMedia.h"
 #include "td/telegram/td_api.h"
 #include "td/telegram/telegram_api.h"
@@ -14,6 +15,7 @@
 
 #include "td/utils/common.h"
 #include "td/utils/FlatHashMap.h"
+#include "td/utils/FlatHashSet.h"
 
 namespace td {
 
@@ -29,10 +31,14 @@ class VoiceNotesManager {
 
   void create_voice_note(FileId file_id, string mime_type, int32 duration, string waveform, bool replace);
 
+  void register_voice_note(FileId voice_note_file_id, FullMessageId full_message_id, const char *source);
+
+  void unregister_voice_note(FileId voice_note_file_id, FullMessageId full_message_id, const char *source);
+
   tl_object_ptr<telegram_api::InputMedia> get_input_media(FileId file_id,
                                                           tl_object_ptr<telegram_api::InputFile> input_file) const;
 
-  SecretInputMedia get_secret_input_media(FileId voice_file_id,
+  SecretInputMedia get_secret_input_media(FileId voice_note_file_id,
                                           tl_object_ptr<telegram_api::InputEncryptedFile> input_file,
                                           const string &caption, int32 layer) const;
 
@@ -65,6 +71,8 @@ class VoiceNotesManager {
 
   Td *td_;
   FlatHashMap<FileId, unique_ptr<VoiceNote>, FileIdHash> voice_notes_;
+
+  FlatHashMap<FileId, FlatHashSet<FullMessageId, FullMessageIdHash>, FileIdHash> voice_note_messages_;
 };
 
 }  // namespace td