Merge 33048ebc01 into 97b29d177a

* Implement renaming in model editing dialog

* add transcription faq

* remove incorrect constraint for viewer as username

should be able to change anyone's role other than admin

* Don't save redundant state changes

* prevent crash when a camera doesn't support onvif imaging service required for focus support

* Fine tune behavior

* Stop redundant go2rtc stream metadata requests and defer audio information to allow bandwidth for image requests

* Improve cleanup logic for capture process

---------

Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>

README_CN.md

@@ -1,28 +1,31 @@
 <p align="center">
-  <img align="center" alt="logo" src="docs/static/img/frigate.png">
+  <img align="center" alt="logo" src="docs/static/img/branding/frigate.png">
 </p>
 
-# Frigate - 一个具有实时目标检测的本地NVR
+# Frigate NVR™ - 一个具有实时目标检测的本地 NVR
 
 [English](https://github.com/blakeblackshear/frigate) | \[简体中文\]
 
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
 <a href="https://hosted.weblate.org/engage/frigate-nvr/-/zh_Hans/">
 <img src="https://hosted.weblate.org/widget/frigate-nvr/-/zh_Hans/svg-badge.svg" alt="翻译状态" />
 </a>
 
-一个完整的本地网络视频录像机（NVR），专为[Home Assistant](https://www.home-assistant.io)设计，具备AI物体检测功能。使用OpenCV和TensorFlow在本地为IP摄像头执行实时物体检测。
+一个完整的本地网络视频录像机（NVR），专为[Home Assistant](https://www.home-assistant.io)设计，具备 AI 目标/物体检测功能。使用 OpenCV 和 TensorFlow 在本地为 IP 摄像头执行实时物体检测。
 
-强烈推荐使用GPU或者AI加速器（例如[Google Coral加速器](https://coral.ai/products/) 或者 [Hailo](https://hailo.ai/)）。它们的性能甚至超过目前的顶级CPU，并且可以以极低的耗电实现更优的性能。
-- 通过[自定义组件](https://github.com/blakeblackshear/frigate-hass-integration)与Home Assistant紧密集成
-- 设计上通过仅在必要时和必要地点寻找物体，最大限度地减少资源使用并最大化性能
+强烈推荐使用 GPU 或者 AI 加速器（例如[Google Coral 加速器](https://coral.ai/products/) 或者 [Hailo](https://hailo.ai/)等）。它们的运行效率远远高于现在的顶级 CPU，并且功耗也极低。
+
+- 通过[自定义组件](https://github.com/blakeblackshear/frigate-hass-integration)与 Home Assistant 紧密集成
+- 设计上通过仅在必要时和必要地点寻找目标，最大限度地减少资源使用并最大化性能
 - 大量利用多进程处理，强调实时性而非处理每一帧
-- 使用非常低开销的运动检测来确定运行物体检测的位置
-- 使用TensorFlow进行物体检测，运行在单独的进程中以达到最大FPS
-- 通过MQTT进行通信，便于集成到其他系统中
+- 使用非常低开销的画面变动检测（也叫运动检测）来确定运行目标检测的位置
+- 使用 TensorFlow 进行目标检测，并运行在单独的进程中以达到最大 FPS
+- 通过 MQTT 进行通信，便于集成到其他系统中
 - 根据检测到的物体设置保留时间进行视频录制
-- 24/7全天候录制
-- 通过RTSP重新流传输以减少摄像头的连接数
-- 支持WebRTC和MSE，实现低延迟的实时观看
+- 24/7 全天候录制
+- 通过 RTSP 重新流传输以减少摄像头的连接数
+- 支持 WebRTC 和 MSE，实现低延迟的实时观看
 
 ## 社区中文翻译文档
 
@@ -32,39 +35,55 @@
 
 如果您想通过捐赠支持开发，请使用 [Github Sponsors](https://github.com/sponsors/blakeblackshear)。
 
+## 协议
+
+本项目采用 **MIT 许可证**授权。
+**代码部分**：本代码库中的源代码、配置文件和文档均遵循 [MIT 许可证](LICENSE)。您可以自由使用、修改和分发这些代码，但必须保留原始版权声明。
+
+**商标部分**：“Frigate”名称、“Frigate NVR”品牌以及 Frigate 的 Logo 为 **Frigate LLC 的商标**，**不在** MIT 许可证覆盖范围内。
+有关品牌资产的规范使用详情，请参阅我们的[《商标政策》](TRADEMARK.md)。
+
 ## 截图
 
 ### 实时监控面板
+
 <div>
 <img width="800" alt="实时监控面板" src="https://github.com/blakeblackshear/frigate/assets/569905/5e713cb9-9db5-41dc-947a-6937c3bc376e">
 </div>
 
 ### 简单的核查工作流程
+
 <div>
 <img width="800" alt="简单的审查工作流程" src="https://github.com/blakeblackshear/frigate/assets/569905/6fed96e8-3b18-40e5-9ddc-31e6f3c9f2ff">
 </div>
 
 ### 多摄像头可按时间轴查看
+
 <div>
 <img width="800" alt="多摄像头可按时间轴查看" src="https://github.com/blakeblackshear/frigate/assets/569905/d6788a15-0eeb-4427-a8d4-80b93cae3d74">
 </div>
 
 ### 内置遮罩和区域编辑器
+
 <div>
 <img width="800" alt="内置遮罩和区域编辑器" src="https://github.com/blakeblackshear/frigate/assets/569905/d7885fc3-bfe6-452f-b7d0-d957cb3e31f5">
 </div>
 
-
 ## 翻译
+
 我们使用 [Weblate](https://hosted.weblate.org/projects/frigate-nvr/) 平台提供翻译支持，欢迎参与进来一起完善。
 
-
 ## 非官方中文讨论社区
-欢迎加入中文讨论QQ群：[1043861059](https://qm.qq.com/q/7vQKsTmSz)
+
+欢迎加入中文讨论 QQ 群：[1043861059](https://qm.qq.com/q/7vQKsTmSz)
 
 Bilibili：https://space.bilibili.com/3546894915602564
 
-
 ## 中文社区赞助商
+
 [![EdgeOne](https://edgeone.ai/media/34fe3a45-492d-4ea4-ae5d-ea1087ca7b4b.png)](https://edgeone.ai/zh?from=github)
 本项目 CDN 加速及安全防护由 Tencent EdgeOne 赞助
+
+---
+
+**Copyright © 2025 Frigate LLC.**

docs/docs/configuration/audio_detectors.md

@@ -157,3 +157,19 @@ Only one `speech` event may be transcribed at a time. Frigate does not automatic
 :::
 
 Recorded `speech` events will always use a `whisper` model, regardless of the `model_size` config setting. Without a supported Nvidia GPU, generating transcriptions for longer `speech` events may take a fair amount of time, so be patient.
+
+#### FAQ
+
+1. Why doesn't Frigate automatically transcribe all `speech` events?
+
+   Frigate does not implement a queue mechanism for speech transcription, and adding one is not trivial. A proper queue would need backpressure, prioritization, memory/disk buffering, retry logic, crash recovery, and safeguards to prevent unbounded growth when events outpace processing. That’s a significant amount of complexity for a feature that, in most real-world environments, would mostly just churn through low-value noise.
+
+   Because transcription is **serialized (one event at a time)** and speech events can be generated far faster than they can be processed, an auto-transcribe toggle would very quickly create an ever-growing backlog and degrade core functionality. For the amount of engineering and risk involved, it adds **very little practical value** for the majority of deployments, which are often on low-powered, edge hardware.
+
+   If you hear speech that’s actually important and worth saving/indexing for the future, **just press the transcribe button in Explore** on that specific `speech` event - that keeps things explicit, reliable, and under your control.
+
+2. Why don't you save live transcription text and use that for `speech` events?
+
+   There’s no guarantee that a `speech` event is even created from the exact audio that went through the transcription model. Live transcription and `speech` event creation are **separate, asynchronous processes**. Even when both are correctly configured, trying to align the **precise start and end time of a speech event** with whatever audio the model happened to be processing at that moment is unreliable.
+
+   Automatically persisting that data would often result in **misaligned, partial, or irrelevant transcripts**, while still incurring all of the CPU, storage, and privacy costs of transcription. That’s why Frigate treats transcription as an **explicit, user-initiated action** rather than an automatic side-effect of every `speech` event.

docs/src/components/LanguageAlert/styles.module.css

@@ -1,13 +1,18 @@
 .alert {
-    padding: 12px;
-    background: #fff8e6;
-    border-bottom: 1px solid #ffd166;
-    text-align: center;
-    font-size: 15px;
-  }
+  padding: 12px;
+  background: #fff8e6;
+  border-bottom: 1px solid #ffd166;
+  text-align: center;
+  font-size: 15px;
+}
 
-  .alert a {
-    color: #1890ff;
-    font-weight: 500;
-    margin-left: 6px;
-  }
+[data-theme="dark"] .alert {
+  background: #3b2f0b;
+  border-bottom: 1px solid #665c22;
+}
+
+.alert a {
+  color: #1890ff;
+  font-weight: 500;
+  margin-left: 6px;
+}

frigate/data_processing/real_time/custom_classification.py

@@ -99,6 +99,42 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         if self.inference_speed:
             self.inference_speed.update(duration)
 
+    def _should_save_image(
+        self, camera: str, detected_state: str, score: float = 1.0
+    ) -> bool:
+        """
+        Determine if we should save the image for training.
+        Save when:
+        - State is changing or being verified (regardless of score)
+        - Score is less than 100% (even if state matches, useful for training)
+        Don't save when:
+        - State is stable (matches current_state) AND score is 100%
+        """
+        if camera not in self.state_history:
+            # First detection for this camera, save it
+            return True
+
+        verification = self.state_history[camera]
+        current_state = verification.get("current_state")
+        pending_state = verification.get("pending_state")
+
+        # Save if there's a pending state change being verified
+        if pending_state is not None:
+            return True
+
+        # Save if the detected state differs from the current verified state
+        # (state is changing)
+        if current_state is not None and detected_state != current_state:
+            return True
+
+        # If score is less than 100%, save even if state matches
+        # (useful for training to improve confidence)
+        if score < 1.0:
+            return True
+
+        # Don't save if state is stable (detected_state == current_state) AND score is 100%
+        return False
+
     def verify_state_change(self, camera: str, detected_state: str) -> str | None:
         """
         Verify state change requires 3 consecutive identical states before publishing.
@@ -212,14 +248,16 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
                 return
 
         if self.interpreter is None:
-            write_classification_attempt(
-                self.train_dir,
-                cv2.cvtColor(frame, cv2.COLOR_RGB2BGR),
-                "none-none",
-                now,
-                "unknown",
-                0.0,
-            )
+            # When interpreter is None, always save (score is 0.0, which is < 1.0)
+            if self._should_save_image(camera, "unknown", 0.0):
+                write_classification_attempt(
+                    self.train_dir,
+                    cv2.cvtColor(frame, cv2.COLOR_RGB2BGR),
+                    "none-none",
+                    now,
+                    "unknown",
+                    0.0,
+                )
             return
 
         input = np.expand_dims(resized_frame, axis=0)
@@ -236,14 +274,17 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         score = round(probs[best_id], 2)
         self.__update_metrics(datetime.datetime.now().timestamp() - now)
 
-        write_classification_attempt(
-            self.train_dir,
-            cv2.cvtColor(frame, cv2.COLOR_RGB2BGR),
-            "none-none",
-            now,
-            self.labelmap[best_id],
-            score,
-        )
+        detected_state = self.labelmap[best_id]
+
+        if self._should_save_image(camera, detected_state, score):
+            write_classification_attempt(
+                self.train_dir,
+                cv2.cvtColor(frame, cv2.COLOR_RGB2BGR),
+                "none-none",
+                now,
+                detected_state,
+                score,
+            )
 
         if score < self.model_config.threshold:
             logger.debug(
@@ -251,7 +292,6 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
             )
             return
 
-        detected_state = self.labelmap[best_id]
         verified_state = self.verify_state_change(camera, detected_state)
 
         if verified_state is not None:

frigate/ptz/onvif.py

@@ -190,7 +190,11 @@ class OnvifController:
         ptz: ONVIFService = await onvif.create_ptz_service()
         self.cams[camera_name]["ptz"] = ptz
 
-        imaging: ONVIFService = await onvif.create_imaging_service()
+        try:
+            imaging: ONVIFService = await onvif.create_imaging_service()
+        except (Fault, ONVIFError, TransportError, Exception) as e:
+            logger.debug(f"Imaging service not supported for {camera_name}: {e}")
+            imaging = None
         self.cams[camera_name]["imaging"] = imaging
         try:
             video_sources = await media.GetVideoSources()
@@ -381,7 +385,10 @@ class OnvifController:
                             f"Disabling autotracking zooming for {camera_name}: Absolute zoom not supported. Exception: {e}"
                         )
 
-        if self.cams[camera_name]["video_source_token"] is not None:
+        if (
+            self.cams[camera_name]["video_source_token"] is not None
+            and imaging is not None
+        ):
             try:
                 imaging_capabilities = await imaging.GetImagingSettings(
                     {"VideoSourceToken": self.cams[camera_name]["video_source_token"]}
@@ -421,6 +428,7 @@ class OnvifController:
         if (
             "focus" in self.cams[camera_name]["features"]
             and self.cams[camera_name]["video_source_token"]
+            and self.cams[camera_name]["imaging"] is not None
         ):
             try:
                 stop_request = self.cams[camera_name]["imaging"].create_type("Stop")
@@ -648,6 +656,7 @@ class OnvifController:
         if (
             "focus" not in self.cams[camera_name]["features"]
             or not self.cams[camera_name]["video_source_token"]
+            or self.cams[camera_name]["imaging"] is None
         ):
             logger.error(f"{camera_name} does not support ONVIF continuous focus.")
             return

frigate/video.py

@@ -124,45 +124,50 @@ def capture_frames(
         config_subscriber.check_for_updates()
         return config.enabled
 
-    while not stop_event.is_set():
-        if not get_enabled_state():
-            logger.debug(f"Stopping capture thread for disabled {config.name}")
-            break
-
-        fps.value = frame_rate.eps()
-        skipped_fps.value = skipped_eps.eps()
-        current_frame.value = datetime.now().timestamp()
-        frame_name = f"{config.name}_frame{frame_index}"
-        frame_buffer = frame_manager.write(frame_name)
-        try:
-            frame_buffer[:] = ffmpeg_process.stdout.read(frame_size)
-        except Exception:
-            # shutdown has been initiated
-            if stop_event.is_set():
+    try:
+        while not stop_event.is_set():
+            if not get_enabled_state():
+                logger.debug(f"Stopping capture thread for disabled {config.name}")
                 break
 
-            logger.error(f"{config.name}: Unable to read frames from ffmpeg process.")
+            fps.value = frame_rate.eps()
+            skipped_fps.value = skipped_eps.eps()
+            current_frame.value = datetime.now().timestamp()
+            frame_name = f"{config.name}_frame{frame_index}"
+            frame_buffer = frame_manager.write(frame_name)
+            try:
+                frame_buffer[:] = ffmpeg_process.stdout.read(frame_size)
+            except Exception:
+                # shutdown has been initiated
+                if stop_event.is_set():
+                    break
 
-            if ffmpeg_process.poll() is not None:
                 logger.error(
-                    f"{config.name}: ffmpeg process is not running. exiting capture thread..."
+                    f"{config.name}: Unable to read frames from ffmpeg process."
                 )
-                break
 
-            continue
+                if ffmpeg_process.poll() is not None:
+                    logger.error(
+                        f"{config.name}: ffmpeg process is not running. exiting capture thread..."
+                    )
+                    break
 
-        frame_rate.update()
+                continue
 
-        # don't lock the queue to check, just try since it should rarely be full
-        try:
-            # add to the queue
-            frame_queue.put((frame_name, current_frame.value), False)
-            frame_manager.close(frame_name)
-        except queue.Full:
-            # if the queue is full, skip this frame
-            skipped_eps.update()
+            frame_rate.update()
 
-        frame_index = 0 if frame_index == shm_frame_count - 1 else frame_index + 1
+            # don't lock the queue to check, just try since it should rarely be full
+            try:
+                # add to the queue
+                frame_queue.put((frame_name, current_frame.value), False)
+                frame_manager.close(frame_name)
+            except queue.Full:
+                # if the queue is full, skip this frame
+                skipped_eps.update()
+
+            frame_index = 0 if frame_index == shm_frame_count - 1 else frame_index + 1
+    finally:
+        config_subscriber.stop()
 
 
 class CameraWatchdog(threading.Thread):
@@ -234,6 +239,16 @@ class CameraWatchdog(threading.Thread):
                 else:
                     self.ffmpeg_detect_process.wait()
 
+        # Wait for old capture thread to fully exit before starting a new one
+        if self.capture_thread is not None and self.capture_thread.is_alive():
+            self.logger.info("Waiting for capture thread to exit...")
+            self.capture_thread.join(timeout=5)
+
+            if self.capture_thread.is_alive():
+                self.logger.warning(
+                    f"Capture thread for {self.config.name} did not exit in time"
+                )
+
         self.logger.error(
             "The following ffmpeg logs include the last 100 lines prior to exit."
         )

web/src/components/classification/ClassificationModelEditDialog.tsx

@@ -37,7 +37,7 @@ import { useForm } from "react-hook-form";
 import { useTranslation } from "react-i18next";
 import { LuPlus, LuX } from "react-icons/lu";
 import { toast } from "sonner";
-import useSWR from "swr";
+import useSWR, { mutate } from "swr";
 import { z } from "zod";
 
 type ClassificationModelEditDialogProps = {
@@ -240,15 +240,61 @@ export default function ClassificationModelEditDialog({
             position: "top-center",
           });
         } else {
-          // State model - update classes
-          // Note: For state models, updating classes requires renaming categories
-          // which is handled through the dataset API, not the config API
-          // We'll need to implement this by calling the rename endpoint for each class
-          // For now, we just show a message that this requires retraining
+          const stateData = data as StateFormData;
+          const newClasses = stateData.classes.filter(
+            (c) => c.trim().length > 0,
+          );
+          const oldClasses = dataset?.categories
+            ? Object.keys(dataset.categories).filter((key) => key !== "none")
+            : [];
 
-          toast.info(t("edit.stateClassesInfo"), {
-            position: "top-center",
-          });
+          const renameMap = new Map<string, string>();
+          const maxLength = Math.max(oldClasses.length, newClasses.length);
+
+          for (let i = 0; i < maxLength; i++) {
+            const oldClass = oldClasses[i];
+            const newClass = newClasses[i];
+
+            if (oldClass && newClass && oldClass !== newClass) {
+              renameMap.set(oldClass, newClass);
+            }
+          }
+
+          const renamePromises = Array.from(renameMap.entries()).map(
+            async ([oldName, newName]) => {
+              try {
+                await axios.put(
+                  `/classification/${model.name}/dataset/${oldName}/rename`,
+                  {
+                    new_category: newName,
+                  },
+                );
+              } catch (err) {
+                const error = err as {
+                  response?: { data?: { message?: string; detail?: string } };
+                };
+                const errorMessage =
+                  error.response?.data?.message ||
+                  error.response?.data?.detail ||
+                  "Unknown error";
+                throw new Error(
+                  `Failed to rename ${oldName} to ${newName}: ${errorMessage}`,
+                );
+              }
+            },
+          );
+
+          if (renamePromises.length > 0) {
+            await Promise.all(renamePromises);
+            await mutate(`classification/${model.name}/dataset`);
+            toast.success(t("toast.success.updatedModel"), {
+              position: "top-center",
+            });
+          } else {
+            toast.info(t("edit.stateClassesInfo"), {
+              position: "top-center",
+            });
+          }
         }
 
         onSuccess();
@@ -256,8 +302,10 @@ export default function ClassificationModelEditDialog({
       } catch (err) {
         const error = err as {
           response?: { data?: { message?: string; detail?: string } };
+          message?: string;
         };
         const errorMessage =
+          error.message ||
           error.response?.data?.message ||
           error.response?.data?.detail ||
           "Unknown error";
@@ -268,7 +316,7 @@ export default function ClassificationModelEditDialog({
         setIsSaving(false);
       }
     },
-    [isObjectModel, model, t, onSuccess, onClose],
+    [isObjectModel, model, dataset, t, onSuccess, onClose],
   );
 
   const handleCancel = useCallback(() => {

web/src/components/menu/LiveContextMenu.tsx

@@ -48,6 +48,7 @@ import { useTranslation } from "react-i18next";
 import { useDateLocale } from "@/hooks/use-date-locale";
 import { useIsAdmin } from "@/hooks/use-is-admin";
 import { CameraNameLabel } from "../camera/FriendlyNameLabel";
+import { LiveStreamMetadata } from "@/types/live";
 
 type LiveContextMenuProps = {
   className?: string;
@@ -68,6 +69,7 @@ type LiveContextMenuProps = {
   resetPreferredLiveMode: () => void;
   config?: FrigateConfig;
   children?: ReactNode;
+  streamMetadata?: { [key: string]: LiveStreamMetadata };
 };
 export default function LiveContextMenu({
   className,
@@ -88,6 +90,7 @@ export default function LiveContextMenu({
   resetPreferredLiveMode,
   config,
   children,
+  streamMetadata,
 }: LiveContextMenuProps) {
   const { t } = useTranslation("views/live");
   const [showSettings, setShowSettings] = useState(false);
@@ -558,6 +561,7 @@ export default function LiveContextMenu({
           setGroupStreamingSettings={setGroupStreamingSettings}
           setIsDialogOpen={setShowSettings}
           onSave={onSave}
+          streamMetadata={streamMetadata}
         />
       </Dialog>
     </div>

web/src/components/settings/CameraStreamingDialog.tsx

@@ -38,6 +38,7 @@ import { useCameraFriendlyName } from "@/hooks/use-camera-friendly-name";
 type CameraStreamingDialogProps = {
   camera: string;
   groupStreamingSettings: GroupStreamingSettings;
+  streamMetadata?: { [key: string]: LiveStreamMetadata };
   setGroupStreamingSettings: React.Dispatch<
     React.SetStateAction<GroupStreamingSettings>
   >;
@@ -48,6 +49,7 @@ type CameraStreamingDialogProps = {
 export function CameraStreamingDialog({
   camera,
   groupStreamingSettings,
+  streamMetadata,
   setGroupStreamingSettings,
   setIsDialogOpen,
   onSave,
@@ -76,12 +78,7 @@ export function CameraStreamingDialog({
     [config, streamName],
   );
 
-  const { data: cameraMetadata } = useSWR<LiveStreamMetadata>(
-    isRestreamed ? `go2rtc/streams/${streamName}` : null,
-    {
-      revalidateOnFocus: false,
-    },
-  );
+  const cameraMetadata = streamName ? streamMetadata?.[streamName] : undefined;
 
   const supportsAudioOutput = useMemo(() => {
     if (!cameraMetadata) {

web/src/hooks/use-camera-live-mode.ts

@@ -1,8 +1,8 @@
-import { baseUrl } from "@/api/baseUrl";
 import { CameraConfig, FrigateConfig } from "@/types/frigateConfig";
 import { useCallback, useEffect, useState, useMemo } from "react";
 import useSWR from "swr";
-import { LivePlayerMode, LiveStreamMetadata } from "@/types/live";
+import { LivePlayerMode } from "@/types/live";
+import useDeferredStreamMetadata from "./use-deferred-stream-metadata";
 
 export default function useCameraLiveMode(
   cameras: CameraConfig[],
@@ -11,9 +11,9 @@ export default function useCameraLiveMode(
 ) {
   const { data: config } = useSWR<FrigateConfig>("config");
 
-  // Get comma-separated list of restreamed stream names for SWR key
-  const restreamedStreamsKey = useMemo(() => {
-    if (!cameras || !config) return null;
+  // Compute which streams need metadata (restreamed streams only)
+  const restreamedStreamNames = useMemo(() => {
+    if (!cameras || !config) return [];
 
     const streamNames = new Set<string>();
     cameras.forEach((camera) => {
@@ -32,56 +32,13 @@ export default function useCameraLiveMode(
       }
     });
 
-    return streamNames.size > 0
-      ? Array.from(streamNames).sort().join(",")
-      : null;
+    return Array.from(streamNames);
   }, [cameras, config, activeStreams]);
 
-  const streamsFetcher = useCallback(async (key: string) => {
-    const streamNames = key.split(",");
-
-    const metadataPromises = streamNames.map(async (streamName) => {
-      try {
-        const response = await fetch(
-          `${baseUrl}api/go2rtc/streams/${streamName}`,
-          {
-            priority: "low",
-          },
-        );
-
-        if (response.ok) {
-          const data = await response.json();
-          return { streamName, data };
-        }
-        return { streamName, data: null };
-      } catch (error) {
-        // eslint-disable-next-line no-console
-        console.error(`Failed to fetch metadata for ${streamName}:`, error);
-        return { streamName, data: null };
-      }
-    });
-
-    const results = await Promise.allSettled(metadataPromises);
-
-    const metadata: { [key: string]: LiveStreamMetadata } = {};
-    results.forEach((result) => {
-      if (result.status === "fulfilled" && result.value.data) {
-        metadata[result.value.streamName] = result.value.data;
-      }
-    });
-
-    return metadata;
-  }, []);
-
-  const { data: allStreamMetadata = {} } = useSWR<{
-    [key: string]: LiveStreamMetadata;
-  }>(restreamedStreamsKey, streamsFetcher, {
-    revalidateOnFocus: false,
-    revalidateOnReconnect: false,
-    revalidateIfStale: false,
-    dedupingInterval: 60000,
-  });
+  // Fetch stream metadata with deferred loading (doesn't block initial render)
+  const streamMetadata = useDeferredStreamMetadata(restreamedStreamNames);
 
+  // Compute live mode states
   const [preferredLiveModes, setPreferredLiveModes] = useState<{
     [key: string]: LivePlayerMode;
   }>({});
@@ -122,10 +79,10 @@ export default function useCameraLiveMode(
         newPreferredLiveModes[camera.name] = isRestreamed ? "mse" : "jsmpeg";
       }
 
-      // check each stream for audio support
+      // Check each stream for audio support
       if (isRestreamed) {
         Object.values(camera.live.streams).forEach((streamName) => {
-          const metadata = allStreamMetadata?.[streamName];
+          const metadata = streamMetadata[streamName];
           newSupportsAudioOutputStates[streamName] = {
             supportsAudio: metadata
               ? metadata.producers.find(
@@ -150,7 +107,7 @@ export default function useCameraLiveMode(
     setPreferredLiveModes(newPreferredLiveModes);
     setIsRestreamedStates(newIsRestreamedStates);
     setSupportsAudioOutputStates(newSupportsAudioOutputStates);
-  }, [cameras, config, windowVisible, allStreamMetadata]);
+  }, [cameras, config, windowVisible, streamMetadata]);
 
   const resetPreferredLiveMode = useCallback(
     (cameraName: string) => {
@@ -180,5 +137,6 @@ export default function useCameraLiveMode(
     resetPreferredLiveMode,
     isRestreamedStates,
     supportsAudioOutputStates,
+    streamMetadata,
   };
 }

web/src/hooks/use-deferred-stream-metadata.ts

@@ -0,0 +1,90 @@
+import { baseUrl } from "@/api/baseUrl";
+import { useCallback, useEffect, useState, useMemo } from "react";
+import useSWR from "swr";
+import { LiveStreamMetadata } from "@/types/live";
+
+const FETCH_TIMEOUT_MS = 10000;
+const DEFER_DELAY_MS = 2000;
+
+/**
+ * Hook that fetches go2rtc stream metadata with deferred loading.
+ *
+ * Metadata fetching is delayed to prevent blocking initial page load
+ * and camera image requests.
+ *
+ * @param streamNames - Array of stream names to fetch metadata for
+ * @returns Object containing stream metadata keyed by stream name
+ */
+export default function useDeferredStreamMetadata(streamNames: string[]) {
+  const [fetchEnabled, setFetchEnabled] = useState(false);
+
+  useEffect(() => {
+    const timeoutId = setTimeout(() => {
+      setFetchEnabled(true);
+    }, DEFER_DELAY_MS);
+
+    return () => clearTimeout(timeoutId);
+  }, []);
+
+  const swrKey = useMemo(() => {
+    if (!fetchEnabled || streamNames.length === 0) return null;
+    // Use spread to avoid mutating the original array
+    return `deferred-streams:${[...streamNames].sort().join(",")}`;
+  }, [fetchEnabled, streamNames]);
+
+  const fetcher = useCallback(async (key: string) => {
+    // Extract stream names from key (remove prefix)
+    const names = key.replace("deferred-streams:", "").split(",");
+
+    const promises = names.map(async (streamName) => {
+      const controller = new AbortController();
+      const timeoutId = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS);
+
+      try {
+        const response = await fetch(
+          `${baseUrl}api/go2rtc/streams/${streamName}`,
+          {
+            priority: "low",
+            signal: controller.signal,
+          },
+        );
+        clearTimeout(timeoutId);
+
+        if (response.ok) {
+          const data = await response.json();
+          return { streamName, data };
+        }
+        return { streamName, data: null };
+      } catch (error) {
+        clearTimeout(timeoutId);
+        if ((error as Error).name !== "AbortError") {
+          // eslint-disable-next-line no-console
+          console.error(`Failed to fetch metadata for ${streamName}:`, error);
+        }
+        return { streamName, data: null };
+      }
+    });
+
+    const results = await Promise.allSettled(promises);
+
+    const metadata: { [key: string]: LiveStreamMetadata } = {};
+    results.forEach((result) => {
+      if (result.status === "fulfilled" && result.value.data) {
+        metadata[result.value.streamName] = result.value.data;
+      }
+    });
+
+    return metadata;
+  }, []);
+
+  const { data: metadata = {} } = useSWR<{
+    [key: string]: LiveStreamMetadata;
+  }>(swrKey, fetcher, {
+    revalidateOnFocus: false,
+    revalidateOnReconnect: false,
+    revalidateIfStale: false,
+    dedupingInterval: 60000,
+  });
+
+  return metadata;
+}

web/src/views/live/DraggableGridLayout.tsx

@@ -24,6 +24,7 @@ import "react-resizable/css/styles.css";
 import {
   AudioState,
   LivePlayerMode,
+  LiveStreamMetadata,
   StatsState,
   VolumeState,
 } from "@/types/live";
@@ -47,7 +48,6 @@ import {
   TooltipContent,
 } from "@/components/ui/tooltip";
 import { Toaster } from "@/components/ui/sonner";
-import useCameraLiveMode from "@/hooks/use-camera-live-mode";
 import LiveContextMenu from "@/components/menu/LiveContextMenu";
 import { useStreamingSettings } from "@/context/streaming-settings-provider";
 import { useTranslation } from "react-i18next";
@@ -65,6 +65,16 @@ type DraggableGridLayoutProps = {
   setIsEditMode: React.Dispatch<React.SetStateAction<boolean>>;
   fullscreen: boolean;
   toggleFullscreen: () => void;
+  preferredLiveModes: { [key: string]: LivePlayerMode };
+  setPreferredLiveModes: React.Dispatch<
+    React.SetStateAction<{ [key: string]: LivePlayerMode }>
+  >;
+  resetPreferredLiveMode: (cameraName: string) => void;
+  isRestreamedStates: { [key: string]: boolean };
+  supportsAudioOutputStates: {
+    [key: string]: { supportsAudio: boolean; cameraName: string };
+  };
+  streamMetadata: { [key: string]: LiveStreamMetadata };
 };
 export default function DraggableGridLayout({
   cameras,
@@ -79,6 +89,12 @@ export default function DraggableGridLayout({
   setIsEditMode,
   fullscreen,
   toggleFullscreen,
+  preferredLiveModes,
+  setPreferredLiveModes,
+  resetPreferredLiveMode,
+  isRestreamedStates,
+  supportsAudioOutputStates,
+  streamMetadata,
 }: DraggableGridLayoutProps) {
   const { t } = useTranslation(["views/live"]);
   const { data: config } = useSWR<FrigateConfig>("config");
@@ -98,33 +114,6 @@ export default function DraggableGridLayout({
     }
   }, [allGroupsStreamingSettings, cameraGroup]);
 
-  const activeStreams = useMemo(() => {
-    const streams: { [cameraName: string]: string } = {};
-    cameras.forEach((camera) => {
-      const availableStreams = camera.live.streams || {};
-      const streamNameFromSettings =
-        currentGroupStreamingSettings?.[camera.name]?.streamName || "";
-      const streamExists =
-        streamNameFromSettings &&
-        Object.values(availableStreams).includes(streamNameFromSettings);
-
-      const streamName = streamExists
-        ? streamNameFromSettings
-        : Object.values(availableStreams)[0] || "";
-
-      streams[camera.name] = streamName;
-    });
-    return streams;
-  }, [cameras, currentGroupStreamingSettings]);
-
-  const {
-    preferredLiveModes,
-    setPreferredLiveModes,
-    resetPreferredLiveMode,
-    isRestreamedStates,
-    supportsAudioOutputStates,
-  } = useCameraLiveMode(cameras, windowVisible, activeStreams);
-
   // grid layout
 
   const ResponsiveGridLayout = useMemo(() => WidthProvider(Responsive), []);
@@ -624,6 +613,7 @@ export default function DraggableGridLayout({
                     resetPreferredLiveMode(camera.name)
                   }
                   config={config}
+                  streamMetadata={streamMetadata}
                 >
                   <LivePlayer
                     key={camera.name}
@@ -838,6 +828,7 @@ type GridLiveContextMenuProps = {
   unmuteAll: () => void;
   resetPreferredLiveMode: () => void;
   config?: FrigateConfig;
+  streamMetadata?: { [key: string]: LiveStreamMetadata };
 };
 
 const GridLiveContextMenu = React.forwardRef<
@@ -868,6 +859,7 @@ const GridLiveContextMenu = React.forwardRef<
       unmuteAll,
       resetPreferredLiveMode,
       config,
+      streamMetadata,
       ...props
     },
     ref,
@@ -899,6 +891,7 @@ const GridLiveContextMenu = React.forwardRef<
           unmuteAll={unmuteAll}
           resetPreferredLiveMode={resetPreferredLiveMode}
           config={config}
+          streamMetadata={streamMetadata}
         >
           {children}
         </LiveContextMenu>

web/src/views/live/LiveDashboardView.tsx

@@ -265,6 +265,7 @@ export default function LiveDashboardView({
     resetPreferredLiveMode,
     isRestreamedStates,
     supportsAudioOutputStates,
+    streamMetadata,
   } = useCameraLiveMode(cameras, windowVisible, activeStreams);
 
   const birdseyeConfig = useMemo(() => config?.birdseye, [config]);
@@ -650,6 +651,12 @@ export default function LiveDashboardView({
               setIsEditMode={setIsEditMode}
               fullscreen={fullscreen}
               toggleFullscreen={toggleFullscreen}
+              preferredLiveModes={preferredLiveModes}
+              setPreferredLiveModes={setPreferredLiveModes}
+              resetPreferredLiveMode={resetPreferredLiveMode}
+              isRestreamedStates={isRestreamedStates}
+              supportsAudioOutputStates={supportsAudioOutputStates}
+              streamMetadata={streamMetadata}
             />
           )}
         </>

web/src/views/settings/AuthenticationView.tsx

@@ -478,33 +478,32 @@ export default function AuthenticationView({
                       <TableCell className="text-right">
                         <TooltipProvider>
                           <div className="flex items-center justify-end gap-2">
-                            {user.username !== "admin" &&
-                              user.username !== "viewer" && (
-                                <Tooltip>
-                                  <TooltipTrigger asChild>
-                                    <Button
-                                      size="sm"
-                                      variant="outline"
-                                      className="h-8 px-2"
-                                      onClick={() => {
-                                        setSelectedUser(user.username);
-                                        setSelectedUserRole(
-                                          user.role || "viewer",
-                                        );
-                                        setShowRoleChange(true);
-                                      }}
-                                    >
-                                      <LuUserCog className="size-3.5" />
-                                      <span className="ml-1.5 hidden sm:inline-block">
-                                        {t("role.title", { ns: "common" })}
-                                      </span>
-                                    </Button>
-                                  </TooltipTrigger>
-                                  <TooltipContent>
-                                    <p>{t("users.table.changeRole")}</p>
-                                  </TooltipContent>
-                                </Tooltip>
-                              )}
+                            {user.username !== "admin" && (
+                              <Tooltip>
+                                <TooltipTrigger asChild>
+                                  <Button
+                                    size="sm"
+                                    variant="outline"
+                                    className="h-8 px-2"
+                                    onClick={() => {
+                                      setSelectedUser(user.username);
+                                      setSelectedUserRole(
+                                        user.role || "viewer",
+                                      );
+                                      setShowRoleChange(true);
+                                    }}
+                                  >
+                                    <LuUserCog className="size-3.5" />
+                                    <span className="ml-1.5 hidden sm:inline-block">
+                                      {t("role.title", { ns: "common" })}
+                                    </span>
+                                  </Button>
+                                </TooltipTrigger>
+                                <TooltipContent>
+                                  <p>{t("users.table.changeRole")}</p>
+                                </TooltipContent>
+                              </Tooltip>
+                            )}
 
                             <Tooltip>
                               <TooltipTrigger asChild>