diff --git a/backend/open_webui/env.py b/backend/open_webui/env.py
index a5f848b62..19e949e88 100644
--- a/backend/open_webui/env.py
+++ b/backend/open_webui/env.py
@@ -311,6 +311,11 @@ RESET_CONFIG_ON_START = (
     os.environ.get("RESET_CONFIG_ON_START", "False").lower() == "true"
 )
 
+
+ENABLE_REALTIME_CHAT_SAVE = (
+    os.environ.get("ENABLE_REALTIME_CHAT_SAVE", "True").lower() == "true"
+)
+
 ####################################
 # REDIS
 ####################################
diff --git a/backend/open_webui/utils/middleware.py b/backend/open_webui/utils/middleware.py
index 7d79e1d0b..a289b2c59 100644
--- a/backend/open_webui/utils/middleware.py
+++ b/backend/open_webui/utils/middleware.py
@@ -65,6 +65,7 @@ from open_webui.env import (
     SRC_LOG_LEVELS,
     GLOBAL_LOG_LEVEL,
     BYPASS_MODEL_ACCESS_CONTROL,
+    ENABLE_REALTIME_CHAT_SAVE,
 )
 from open_webui.constants import TASKS
 
@@ -977,7 +978,6 @@ async def process_chat_response(
                             )
 
                         else:
-
                             value = (
                                 data.get("choices", [])[0]
                                 .get("delta", {})
@@ -987,6 +987,28 @@ async def process_chat_response(
                             if value:
                                 content = f"{content}{value}"
 
+                                if ENABLE_REALTIME_CHAT_SAVE:
+                                    # Save message in the database
+                                    Chats.upsert_message_to_chat_by_id_and_message_id(
+                                        metadata["chat_id"],
+                                        metadata["message_id"],
+                                        {
+                                            "content": content,
+                                        },
+                                    )
+                                else:
+                                    data = {
+                                        "content": content,
+                                    }
+
+                    except Exception as e:
+                        done = "data: [DONE]" in line
+                        title = Chats.get_chat_title_by_id(metadata["chat_id"])
+
+                        if done:
+                            data = {"done": True, "content": content, "title": title}
+
+                            if not ENABLE_REALTIME_CHAT_SAVE:
                                 # Save message in the database
                                 Chats.upsert_message_to_chat_by_id_and_message_id(
                                     metadata["chat_id"],
@@ -996,13 +1018,6 @@ async def process_chat_response(
                                     },
                                 )
 
-                    except Exception as e:
-                        done = "data: [DONE]" in line
-                        title = Chats.get_chat_title_by_id(metadata["chat_id"])
-
-                        if done:
-                            data = {"done": True, "content": content, "title": title}
-
                             # Send a webhook notification if the user is not active
                             if (
                                 get_user_id_from_session_pool(metadata["session_id"])
diff --git a/src/lib/components/chat/Chat.svelte b/src/lib/components/chat/Chat.svelte
index 9bec3934d..6c3d610ad 100644
--- a/src/lib/components/chat/Chat.svelte
+++ b/src/lib/components/chat/Chat.svelte
@@ -1053,7 +1053,7 @@
 	};
 
 	const chatCompletionEventHandler = async (data, message, chatId) => {
-		const { id, done, choices, sources, selected_model_id, error, usage } = data;
+		const { id, done, choices, content, sources, selected_model_id, error, usage } = data;
 
 		if (error) {
 			await handleOpenAIError(error, message);
@@ -1105,6 +1105,38 @@
 			}
 		}
 
+		if (content) {
+			// REALTIME_CHAT_SAVE is disabled
+			message.content = content;
+
+			if (navigator.vibrate && ($settings?.hapticFeedback ?? false)) {
+				navigator.vibrate(5);
+			}
+
+			// Emit chat event for TTS
+			const messageContentParts = getMessageContentParts(
+				message.content,
+				$config?.audio?.tts?.split_on ?? 'punctuation'
+			);
+			messageContentParts.pop();
+
+			// dispatch only last sentence and make sure it hasn't been dispatched before
+			if (
+				messageContentParts.length > 0 &&
+				messageContentParts[messageContentParts.length - 1] !== message.lastSentence
+			) {
+				message.lastSentence = messageContentParts[messageContentParts.length - 1];
+				eventTarget.dispatchEvent(
+					new CustomEvent('chat', {
+						detail: {
+							id: message.id,
+							content: messageContentParts[messageContentParts.length - 1]
+						}
+					})
+				);
+			}
+		}
+
 		if (selected_model_id) {
 			message.selectedModelId = selected_model_id;
 			message.arena = true;