chore: lint fix

* spacing tweaks and add link to explore for plate

* clear selected objects when changing cameras

* plate link and spacing in object lifecycle

* set tabindex to prevent tooltip from showing on reopen

* show month and day in object lifecycle timestamp

docker/main/requirements.txt

@@ -1,2 +1 @@
 scikit-build == 0.18.*
-nvidia-pyindex

docs/docs/configuration/camera_specific.md

@@ -164,13 +164,35 @@ According to [this discussion](https://github.com/blakeblackshear/frigate/issues
 Cameras connected via a Reolink NVR can be connected with the http stream, use `channel[0..15]` in the stream url for the additional channels.
 The setup of main stream can be also done via RTSP, but isn't always reliable on all hardware versions. The example configuration is working with the oldest HW version RLN16-410 device with multiple types of cameras.
 
+<details>
+  <summary>Example Config</summary>
+
+:::tip
+
+Reolink's latest cameras support two way audio via go2rtc and other applications. It is important that the http-flv stream is still used for stability, a secondary rtsp stream can be added that will be using for the two way audio only.
+
+NOTE: The RTSP stream can not be prefixed with `ffmpeg:`, as go2rtc needs to handle the stream to support two way audio.
+
+Ensure HTTP is enabled in the camera's advanced network settings. To use two way talk with Frigate, see the [Live view documentation](/configuration/live#two-way-talk).
+
+:::
+
 ```yaml
 go2rtc:
   streams:
+    # example for connecting to a standard Reolink camera
     your_reolink_camera:
       - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=username&password=password#video=copy#audio=copy#audio=opus"
     your_reolink_camera_sub:
       - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=username&password=password"
+    # example for connectin to a Reolink camera that supports two way talk
+    your_reolink_camera_twt:
+      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=username&password=password#video=copy#audio=copy#audio=opus"
+      - "rtsp://username:password@reolink_ip/Preview_01_sub
+    your_reolink_camera_twt_sub:
+      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=username&password=password"
+      - "rtsp://username:password@reolink_ip/Preview_01_sub
+    # example for connecting to a Reolink NVR
     your_reolink_camera_via_nvr:
       - "ffmpeg:http://reolink_nvr_ip/flv?port=1935&app=bcs&stream=channel3_main.bcs&user=username&password=password" # channel numbers are 0-15
       - "ffmpeg:your_reolink_camera_via_nvr#audio=aac"
@@ -201,22 +223,7 @@ cameras:
           roles:
             - detect
 ```
-
-#### Reolink Doorbell
-
-The reolink doorbell supports two way audio via go2rtc and other applications. It is important that the http-flv stream is still used for stability, a secondary rtsp stream can be added that will be using for the two way audio only.
-
-Ensure HTTP is enabled in the camera's advanced network settings. To use two way talk with Frigate, see the [Live view documentation](/configuration/live#two-way-talk).
-
-```yaml
-go2rtc:
-  streams:
-    your_reolink_doorbell:
-      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=username&password=password#video=copy#audio=copy#audio=opus"
-      - rtsp://reolink_ip/Preview_01_sub
-    your_reolink_doorbell_sub:
-      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=username&password=password"
-```
+</details>
 
 ### Unifi Protect Cameras
 

docs/docs/configuration/face_recognition.md

@@ -161,6 +161,8 @@ Start with the [Usage](#usage) section and re-read the [Model Requirements](#mod
 
 Accuracy is definitely a going to be improved with higher quality cameras / streams. It is important to look at the DORI (Detection Observation Recognition Identification) range of your camera, if that specification is posted. This specification explains the distance from the camera that a person can be detected, observed, recognized, and identified. The identification range is the most relevant here, and the distance listed by the camera is the furthest that face recognition will realistically work.
 
+Some users have also noted that setting the stream in camera firmware to a constant bit rate (CBR) leads to better image clarity than with a variable bit rate (VBR).
+
 ### Why can't I bulk upload photos?
 
 It is important to methodically add photos to the library, bulk importing photos (especially from a general photo library) will lead to over-fitting in that particular scenario and hurt recognition performance.

docs/docs/configuration/genai.md

@@ -17,18 +17,17 @@ To use Generative AI, you must define a single provider at the global level of y
 genai:
   provider: gemini
   api_key: "{FRIGATE_GEMINI_API_KEY}"
-  model: gemini-1.5-flash
+  model: gemini-2.0-flash
 
 cameras:
   front_camera:
-    objects:
     genai:
-        enabled: True # <- enable GenAI for your front camera
-        use_snapshot: True
-        objects:
-          - person
-        required_zones:
-          - steps
+      enabled: True # <- enable GenAI for your front camera
+      use_snapshot: True
+      objects:
+        - person
+      required_zones:
+        - steps
   indoor_camera:
     objects:
       genai:
@@ -80,7 +79,7 @@ Google Gemini has a free tier allowing [15 queries per minute](https://ai.google
 
 ### Supported Models
 
-You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://ai.google.dev/gemini-api/docs/models/gemini). At the time of writing, this includes `gemini-1.5-pro` and `gemini-1.5-flash`.
+You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://ai.google.dev/gemini-api/docs/models/gemini).
 
 ### Get API Key
 
@@ -97,7 +96,7 @@ To start using Gemini, you must first get an API key from [Google AI Studio](htt
 genai:
   provider: gemini
   api_key: "{FRIGATE_GEMINI_API_KEY}"
-  model: gemini-1.5-flash
+  model: gemini-2.0-flash
 ```
 
 :::note
@@ -112,7 +111,7 @@ OpenAI does not have a free tier for their API. With the release of gpt-4o, pric
 
 ### Supported Models
 
-You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://platform.openai.com/docs/models). At the time of writing, this includes `gpt-4o` and `gpt-4-turbo`.
+You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://platform.openai.com/docs/models).
 
 ### Get API Key
 
@@ -139,18 +138,19 @@ Microsoft offers several vision models through Azure OpenAI. A subscription is r
 
 ### Supported Models
 
-You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models). At the time of writing, this includes `gpt-4o` and `gpt-4-turbo`.
+You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models).
 
 ### Create Resource and Get API Key
 
-To start using Azure OpenAI, you must first [create a resource](https://learn.microsoft.com/azure/cognitive-services/openai/how-to/create-resource?pivots=web-portal#create-a-resource). You'll need your API key and resource URL, which must include the `api-version` parameter (see the example below). The model field is not required in your configuration as the model is part of the deployment name you chose when deploying the resource.
+To start using Azure OpenAI, you must first [create a resource](https://learn.microsoft.com/azure/cognitive-services/openai/how-to/create-resource?pivots=web-portal#create-a-resource). You'll need your API key, model name, and resource URL, which must include the `api-version` parameter (see the example below).
 
 ### Configuration
 
 ```yaml
 genai:
   provider: azure_openai
-  base_url: https://example-endpoint.openai.azure.com/openai/deployments/gpt-4o/chat/completions?api-version=2023-03-15-preview
+  base_url: https://instance.cognitiveservices.azure.com/openai/responses?api-version=2025-04-01-preview
+  model: gpt-5-mini
   api_key: "{FRIGATE_OPENAI_API_KEY}"
 ```
 
@@ -196,10 +196,10 @@ genai:
   model: llava
 
 objects:
-    prompt: "Analyze the {label} in these images from the {camera} security camera. Focus on the actions, behavior, and potential intent of the {label}, rather than just describing its appearance."
-    object_prompts:
-      person: "Examine the main person in these images. What are they doing and what might their actions suggest about their intent (e.g., approaching a door, leaving an area, standing still)? Do not describe the surroundings or static details."
-      car: "Observe the primary vehicle in these images. Focus on its movement, direction, or purpose (e.g., parking, approaching, circling). If it's a delivery vehicle, mention the company."
+  prompt: "Analyze the {label} in these images from the {camera} security camera. Focus on the actions, behavior, and potential intent of the {label}, rather than just describing its appearance."
+  object_prompts:
+    person: "Examine the main person in these images. What are they doing and what might their actions suggest about their intent (e.g., approaching a door, leaving an area, standing still)? Do not describe the surroundings or static details."
+    car: "Observe the primary vehicle in these images. Focus on its movement, direction, or purpose (e.g., parking, approaching, circling). If it's a delivery vehicle, mention the company."
 ```
 
 Prompts can also be overridden at the camera level to provide a more detailed prompt to the model about your specific camera, if you desire.

docs/docs/configuration/license_plate_recognition.md

@@ -30,8 +30,7 @@ In the default mode, Frigate's LPR needs to first detect a `car` or `motorcycle`
 
 ## Minimum System Requirements
 
-License plate recognition works by running AI models locally on your system. The models are relatively lightweight and can run on your CPU or GPU, depending on your configuration. At least 4GB of RAM is required.
-
+License plate recognition works by running AI models locally on your system. The YOLOv9 plate detector model and the OCR models ([PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR)) are relatively lightweight and can run on your CPU or GPU, depending on your configuration. At least 4GB of RAM is required.
 ## Configuration
 
 License plate recognition is disabled by default. Enable it in your config file:

docs/docs/configuration/live.md

@@ -174,7 +174,7 @@ For devices that support two way talk, Frigate can be configured to use the feat
 - Ensure you access Frigate via https (may require [opening port 8971](/frigate/installation/#ports)).
 - For the Home Assistant Frigate card, [follow the docs](http://card.camera/#/usage/2-way-audio) for the correct source.
 
-To use the Reolink Doorbell with two way talk, you should use the [recommended Reolink configuration](/configuration/camera_specific#reolink-doorbell)
+To use the Reolink Doorbell with two way talk, you should use the [recommended Reolink configuration](/configuration/camera_specific#reolink-cameras)
 
 As a starting point to check compatibility for your camera, view the list of cameras supported for two-way talk on the [go2rtc repository](https://github.com/AlexxIT/go2rtc?tab=readme-ov-file#two-way-audio). For cameras in the category `ONVIF Profile T`, you can use the [ONVIF Conformant Products Database](https://www.onvif.org/conformant-products/)'s FeatureList to check for the presence of `AudioOutput`. A camera that supports `ONVIF Profile T` _usually_ supports this, but due to inconsistent support, a camera that explicitly lists this feature may still not work. If no entry for your camera exists on the database, it is recommended not to buy it or to consult with the manufacturer's support on the feature availability.
 

docs/docs/configuration/object_detectors.md

@@ -1455,7 +1455,7 @@ COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
 WORKDIR /dfine
 RUN git clone https://github.com/Peterande/D-FINE.git .
 RUN uv pip install --system -r requirements.txt
-RUN uv pip install --system onnx onnxruntime onnxsim
+RUN uv pip install --system onnx onnxruntime onnxsim onnxscript
 # Create output directory and download checkpoint
 RUN mkdir -p output
 ARG MODEL_SIZE
@@ -1479,9 +1479,9 @@ FROM python:3.11 AS build
 RUN apt-get update && apt-get install --no-install-recommends -y libgl1 && rm -rf /var/lib/apt/lists/*
 COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
 WORKDIR /rfdetr
-RUN uv pip install --system rfdetr onnx onnxruntime onnxsim onnx-graphsurgeon
+RUN uv pip install --system rfdetr[onnxexport] torch==2.8.0 onnxscript
 ARG MODEL_SIZE
-RUN python3 -c "from rfdetr import RFDETR${MODEL_SIZE}; x = RFDETR${MODEL_SIZE}(resolution=320); x.export()"
+RUN python3 -c "from rfdetr import RFDETR${MODEL_SIZE}; x = RFDETR${MODEL_SIZE}(resolution=320); x.export(simplify=True)"
 FROM scratch
 ARG MODEL_SIZE
 COPY --from=build /rfdetr/output/inference_model.onnx /rfdetr-${MODEL_SIZE}.onnx
@@ -1529,7 +1529,7 @@ COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
 WORKDIR /yolov9
 ADD https://github.com/WongKinYiu/yolov9.git .
 RUN uv pip install --system -r requirements.txt
-RUN uv pip install --system onnx==1.18.0 onnxruntime onnx-simplifier>=0.4.1
+RUN uv pip install --system onnx==1.18.0 onnxruntime onnx-simplifier>=0.4.1 onnxscript
 ARG MODEL_SIZE
 ARG IMG_SIZE
 ADD https://github.com/WongKinYiu/yolov9/releases/download/v0.1/yolov9-${MODEL_SIZE}-converted.pt yolov9-${MODEL_SIZE}.pt

docs/docs/frigate/updating.md

@@ -5,7 +5,7 @@ title: Updating
 
 # Updating Frigate
 
-The current stable version of Frigate is **0.16.1**. The release notes and any breaking changes for this version can be found on the [Frigate GitHub releases page](https://github.com/blakeblackshear/frigate/releases/tag/v0.16.1).
+The current stable version of Frigate is **0.16.2**. The release notes and any breaking changes for this version can be found on the [Frigate GitHub releases page](https://github.com/blakeblackshear/frigate/releases/tag/v0.16.2).
 
 Keeping Frigate up to date ensures you benefit from the latest features, performance improvements, and bug fixes. The update process varies slightly depending on your installation method (Docker, Home Assistant Addon, etc.). Below are instructions for the most common setups.
 
@@ -33,21 +33,21 @@ If you’re running Frigate via Docker (recommended method), follow these steps:
 2. **Update and Pull the Latest Image**:
 
    - If using Docker Compose:
-     - Edit your `docker-compose.yml` file to specify the desired version tag (e.g., `0.16.1` instead of `0.15.2`). For example:
+     - Edit your `docker-compose.yml` file to specify the desired version tag (e.g., `0.16.2` instead of `0.15.2`). For example:
        ```yaml
        services:
          frigate:
-           image: ghcr.io/blakeblackshear/frigate:0.16.1
+           image: ghcr.io/blakeblackshear/frigate:0.16.2
        ```
      - Then pull the image:
        ```bash
-       docker pull ghcr.io/blakeblackshear/frigate:0.16.1
+       docker pull ghcr.io/blakeblackshear/frigate:0.16.2
        ```
      - **Note for `stable` Tag Users**: If your `docker-compose.yml` uses the `stable` tag (e.g., `ghcr.io/blakeblackshear/frigate:stable`), you don’t need to update the tag manually. The `stable` tag always points to the latest stable release after pulling.
    - If using `docker run`:
-     - Pull the image with the appropriate tag (e.g., `0.16.1`, `0.16.1-tensorrt`, or `stable`):
+     - Pull the image with the appropriate tag (e.g., `0.16.2`, `0.16.2-tensorrt`, or `stable`):
        ```bash
-       docker pull ghcr.io/blakeblackshear/frigate:0.16.1
+       docker pull ghcr.io/blakeblackshear/frigate:0.16.2
        ```
 
 3. **Start the Container**:

docs/docs/integrations/mqtt.md

@@ -161,7 +161,14 @@ Message published for updates to tracked object metadata, for example:
 
 ### `frigate/reviews`
 
-Message published for each changed review item. The first message is published when the `detection` or `alert` is initiated. When additional objects are detected or when a zone change occurs, it will publish a, `update` message with the same id. When the review activity has ended a final `end` message is published.
+Message published for each changed review item. The first message is published when the `detection` or `alert` is initiated. 
+
+An `update` with the same ID will be published when:
+- The severity changes from `detection` to `alert`
+- Additional objects are detected
+- An object is recognized via face, lpr, etc.
+
+When the review activity has ended a final `end` message is published.
 
 ```json
 {

docs/docs/plus/annotating.md

@@ -42,6 +42,7 @@ Misidentified objects should have a correct label added. For example, if a perso
 | `w`               | Add box                       |
 | `d`               | Toggle difficult              |
 | `s`               | Switch to the next label      |
+| `Shift + s`       | Switch to the previous label  |
 | `tab`             | Select next largest box       |
 | `del`             | Delete current box            |
 | `esc`             | Deselect/Cancel               |

frigate/api/app.py

@@ -696,7 +696,11 @@ def timeline(camera: str = "all", limit: int = 100, source_id: Optional[str] = N
         clauses.append((Timeline.camera == camera))
 
     if source_id:
-        clauses.append((Timeline.source_id == source_id))
+        source_ids = [sid.strip() for sid in source_id.split(",")]
+        if len(source_ids) == 1:
+            clauses.append((Timeline.source_id == source_ids[0]))
+        else:
+            clauses.append((Timeline.source_id.in_(source_ids)))
 
     if len(clauses) == 0:
         clauses.append((True))

frigate/api/export.py

@@ -9,6 +9,7 @@ from typing import List
 import psutil
 from fastapi import APIRouter, Depends, Request
 from fastapi.responses import JSONResponse
+from pathvalidate import sanitize_filepath
 from peewee import DoesNotExist
 from playhouse.shortcuts import model_to_dict
 
@@ -26,7 +27,7 @@ from frigate.api.defs.response.export_response import (
 )
 from frigate.api.defs.response.generic_response import GenericResponse
 from frigate.api.defs.tags import Tags
-from frigate.const import EXPORT_DIR
+from frigate.const import CLIPS_DIR, EXPORT_DIR
 from frigate.models import Export, Previews, Recordings
 from frigate.record.export import (
     PlaybackFactorEnum,
@@ -88,7 +89,14 @@ def export_recording(
     playback_factor = body.playback
     playback_source = body.source
     friendly_name = body.name
-    existing_image = body.image_path
+    existing_image = sanitize_filepath(body.image_path) if body.image_path else None
+
+    # Ensure that existing_image is a valid path
+    if existing_image and not existing_image.startswith(CLIPS_DIR):
+        return JSONResponse(
+            content=({"success": False, "message": "Invalid image path"}),
+            status_code=400,
+        )
 
     if playback_source == "recordings":
         recordings_count = (

frigate/data_processing/real_time/custom_classification.py

@@ -53,6 +53,7 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         self.tensor_output_details: dict[str, Any] | None = None
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
+        self.state_history: dict[str, dict[str, Any]] = {}
 
         if (
             self.metrics
@@ -94,6 +95,42 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         if self.inference_speed:
             self.inference_speed.update(duration)
 
+    def verify_state_change(self, camera: str, detected_state: str) -> str | None:
+        """
+        Verify state change requires 3 consecutive identical states before publishing.
+        Returns state to publish or None if verification not complete.
+        """
+        if camera not in self.state_history:
+            self.state_history[camera] = {
+                "current_state": None,
+                "pending_state": None,
+                "consecutive_count": 0,
+            }
+
+        verification = self.state_history[camera]
+
+        if detected_state == verification["current_state"]:
+            verification["pending_state"] = None
+            verification["consecutive_count"] = 0
+            return None
+
+        if detected_state == verification["pending_state"]:
+            verification["consecutive_count"] += 1
+
+            if verification["consecutive_count"] >= 3:
+                verification["current_state"] = detected_state
+                verification["pending_state"] = None
+                verification["consecutive_count"] = 0
+                return detected_state
+        else:
+            verification["pending_state"] = detected_state
+            verification["consecutive_count"] = 1
+            logger.debug(
+                f"New state '{detected_state}' detected for {camera}, need {3 - verification['consecutive_count']} more consecutive detections"
+            )
+
+        return None
+
     def process_frame(self, frame_data: dict[str, Any], frame: np.ndarray):
         if self.metrics and self.model_config.name in self.metrics.classification_cps:
             self.metrics.classification_cps[
@@ -131,6 +168,19 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
                 self.last_run = now
                 should_run = True
 
+        # Shortcut: always run if we have a pending state verification to complete
+        if (
+            not should_run
+            and camera in self.state_history
+            and self.state_history[camera]["pending_state"] is not None
+            and now > self.last_run + 0.5
+        ):
+            self.last_run = now
+            should_run = True
+            logger.debug(
+                f"Running verification check for pending state: {self.state_history[camera]['pending_state']} ({self.state_history[camera]['consecutive_count']}/3)"
+            )
+
         if not should_run:
             return
 
@@ -188,10 +238,19 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
             score,
         )
 
-        if score >= self.model_config.threshold:
+        if score < self.model_config.threshold:
+            logger.debug(
+                f"Score {score} below threshold {self.model_config.threshold}, skipping verification"
+            )
+            return
+
+        detected_state = self.labelmap[best_id]
+        verified_state = self.verify_state_change(camera, detected_state)
+
+        if verified_state is not None:
             self.requestor.send_data(
                 f"{camera}/classification/{self.model_config.name}",
-                self.labelmap[best_id],
+                verified_state,
             )
 
     def handle_request(self, topic, request_data):
@@ -230,7 +289,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         self.sub_label_publisher = sub_label_publisher
         self.tensor_input_details: dict[str, Any] | None = None
         self.tensor_output_details: dict[str, Any] | None = None
-        self.detected_objects: dict[str, float] = {}
+        self.classification_history: dict[str, list[tuple[str, float, float]]] = {}
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
 
@@ -272,6 +331,56 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         if self.inference_speed:
             self.inference_speed.update(duration)
 
+    def get_weighted_score(
+        self,
+        object_id: str,
+        current_label: str,
+        current_score: float,
+        current_time: float,
+    ) -> tuple[str | None, float]:
+        """
+        Determine weighted score based on history to prevent false positives/negatives.
+        Requires 60% of attempts to agree on a label before publishing.
+        Returns (weighted_label, weighted_score) or (None, 0.0) if no weighted score.
+        """
+        if object_id not in self.classification_history:
+            self.classification_history[object_id] = []
+
+        self.classification_history[object_id].append(
+            (current_label, current_score, current_time)
+        )
+
+        history = self.classification_history[object_id]
+
+        if len(history) < 3:
+            return None, 0.0
+
+        label_counts = {}
+        label_scores = {}
+        total_attempts = len(history)
+
+        for label, score, timestamp in history:
+            if label not in label_counts:
+                label_counts[label] = 0
+                label_scores[label] = []
+
+            label_counts[label] += 1
+            label_scores[label].append(score)
+
+        best_label = max(label_counts, key=label_counts.get)
+        best_count = label_counts[best_label]
+
+        consensus_threshold = total_attempts * 0.6
+        if best_count < consensus_threshold:
+            return None, 0.0
+
+        avg_score = sum(label_scores[best_label]) / len(label_scores[best_label])
+
+        if best_label == "none":
+            return None, 0.0
+
+        return best_label, avg_score
+
     def process_frame(self, obj_data, frame):
         if self.metrics and self.model_config.name in self.metrics.classification_cps:
             self.metrics.classification_cps[
@@ -284,6 +393,9 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         if obj_data["label"] not in self.model_config.object_config.objects:
             return
 
+        if obj_data.get("end_time") is not None:
+            return
+
         now = datetime.datetime.now().timestamp()
         x, y, x2, y2 = calculate_region(
             frame.shape,
@@ -331,7 +443,6 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         probs = res / res.sum(axis=0)
         best_id = np.argmax(probs)
         score = round(probs[best_id], 2)
-        previous_score = self.detected_objects.get(obj_data["id"], 0.0)
         self.__update_metrics(datetime.datetime.now().timestamp() - now)
 
         write_classification_attempt(
@@ -347,30 +458,34 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             logger.debug(f"Score {score} is less than threshold.")
             return
 
-        if score <= previous_score:
-            logger.debug(f"Score {score} is worse than previous score {previous_score}")
-            return
-
         sub_label = self.labelmap[best_id]
-        self.detected_objects[obj_data["id"]] = score
 
-        if (
-            self.model_config.object_config.classification_type
-            == ObjectClassificationType.sub_label
-        ):
-            if sub_label != "none":
+        consensus_label, consensus_score = self.get_weighted_score(
+            obj_data["id"], sub_label, score, now
+        )
+
+        if consensus_label is not None:
+            if (
+                self.model_config.object_config.classification_type
+                == ObjectClassificationType.sub_label
+            ):
                 self.sub_label_publisher.publish(
-                    (obj_data["id"], sub_label, score),
+                    (obj_data["id"], consensus_label, consensus_score),
                     EventMetadataTypeEnum.sub_label,
                 )
-        elif (
-            self.model_config.object_config.classification_type
-            == ObjectClassificationType.attribute
-        ):
-            self.sub_label_publisher.publish(
-                (obj_data["id"], self.model_config.name, sub_label, score),
-                EventMetadataTypeEnum.attribute.value,
-            )
+            elif (
+                self.model_config.object_config.classification_type
+                == ObjectClassificationType.attribute
+            ):
+                self.sub_label_publisher.publish(
+                    (
+                        obj_data["id"],
+                        self.model_config.name,
+                        consensus_label,
+                        consensus_score,
+                    ),
+                    EventMetadataTypeEnum.attribute.value,
+                )
 
     def handle_request(self, topic, request_data):
         if topic == EmbeddingsRequestEnum.reload_classification_model.value:
@@ -388,8 +503,8 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             return None
 
     def expire_object(self, object_id, camera):
-        if object_id in self.detected_objects:
-            self.detected_objects.pop(object_id)
+        if object_id in self.classification_history:
+            self.classification_history.pop(object_id)
 
 
 @staticmethod

frigate/genai/__init__.py

@@ -63,18 +63,24 @@ class GenAIClient:
             else:
                 return ""
 
-        def get_verified_objects() -> str:
+        def get_verified_object_prompt() -> str:
             if review_data["recognized_objects"]:
-                return "  - " + "\n  - ".join(review_data["recognized_objects"])
+                object_list = "  - " + "\n  - ".join(review_data["recognized_objects"])
+                return f"""## Verified Objects (USE THESE NAMES)
+When any of the following verified objects are present in the scene, you MUST use these exact names in your title and scene description:
+{object_list}
+"""
             else:
-                return "  None"
+                return ""
 
         context_prompt = f"""
-Please analyze the sequence of images ({len(thumbnails)} total) taken in chronological order from the perspective of the {review_data["camera"].replace("_", " ")} security camera.
+Your task is to analyze the sequence of images ({len(thumbnails)} total) taken in chronological order from the perspective of the {review_data["camera"].replace("_", " ")} security camera.
 
-**Normal activity patterns for this property:**
+## Normal Activity Patterns for This Property
 {activity_context_prompt}
 
+## Task Instructions
+
 Your task is to provide a clear, accurate description of the scene that:
 1. States exactly what is happening based on observable actions and movements.
 2. Evaluates whether the observable evidence suggests normal activity for this property or genuine security concerns.
@@ -82,6 +88,8 @@ Your task is to provide a clear, accurate description of the scene that:
 
 **IMPORTANT: Start by checking if the activity matches the normal patterns above. If it does, assign Level 0. Only consider higher threat levels if the activity clearly deviates from normal patterns or shows genuine security concerns.**
 
+## Analysis Guidelines
+
 When forming your description:
 - **CRITICAL: Only describe objects explicitly listed in "Detected objects" below.** Do not infer or mention additional people, vehicles, or objects not present in the detected objects list, even if visual patterns suggest them. If only a car is detected, do not describe a person interacting with it unless "person" is also in the detected objects list.
 - **Only describe actions actually visible in the frames.** Do not assume or infer actions that you don't observe happening. If someone walks toward furniture but you never see them sit, do not say they sat. Stick to what you can see across the sequence.
@@ -92,6 +100,8 @@ When forming your description:
 - Identify patterns that suggest genuine security concerns: testing doors/windows on vehicles or buildings, accessing unauthorized areas, attempting to conceal actions, extended loitering without apparent purpose, taking items, behavior that clearly doesn't align with the zone context and detected objects.
 - **Weigh all evidence holistically**: Start by checking if the activity matches the normal patterns above. If it does, assign Level 0. Only consider Level 1 if the activity clearly deviates from normal patterns or shows genuine security concerns that warrant attention.
 
+## Response Format
+
 Your response MUST be a flat JSON object with:
 - `title` (string): A concise, one-sentence title that captures the main activity. Include any verified recognized objects (from the "Verified recognized objects" list below) and key detected objects. Examples: "Joe walking dog in backyard", "Unknown person testing car doors at night".
 - `scene` (string): A narrative description of what happens across the sequence from start to finish. **Only describe actions you can actually observe happening in the frames provided.** Do not infer or assume actions that aren't visible (e.g., if you see someone walking but never see them sit, don't say they sat down). Include setting, detected objects, and their observable actions. Avoid speculation or filling in assumed behaviors. Your description should align with and support the threat level you assign.
@@ -99,20 +109,22 @@ Your response MUST be a flat JSON object with:
 - `potential_threat_level` (integer): 0, 1, or 2 as defined below. Your threat level must be consistent with your scene description and the guidance above.
 {get_concern_prompt()}
 
-Threat-level definitions:
+## Threat Level Definitions
+
 - 0 — **Normal activity (DEFAULT)**: What you observe matches the normal activity patterns above or is consistent with expected activity for this property type. The observable evidence—considering zone context, detected objects, and timing together—supports a benign explanation. **Use this level for routine activities even if minor ambiguous elements exist.**
 - 1 — **Potentially suspicious**: Observable behavior raises genuine security concerns that warrant human review. The evidence doesn't support a routine explanation and clearly deviates from the normal patterns above. Examples: testing doors/windows on vehicles or structures, accessing areas that don't align with the activity, taking items that likely don't belong to them, behavior clearly inconsistent with the zone and context, or activity that lacks any visible legitimate indicators. **Only use this level when the activity clearly doesn't match normal patterns.**
 - 2 — **Immediate threat**: Clear evidence of forced entry, break-in, vandalism, aggression, weapons, theft in progress, or active property damage.
 
-Sequence details:
+## Sequence Details
+
 - Frame 1 = earliest, Frame {len(thumbnails)} = latest
 - Activity started at {review_data["start"]} and lasted {review_data["duration"]} seconds
 - Detected objects: {", ".join(review_data["objects"])}
-- Verified recognized objects (use these names when describing these objects):
-{get_verified_objects()}
 - Zones involved: {", ".join(z.replace("_", " ").title() for z in review_data["zones"]) or "None"}
 
-**IMPORTANT:**
+{get_verified_object_prompt()}
+
+## Important Notes
 - Values must be plain strings, floats, or integers — no nested objects, no extra commentary.
 - Only describe objects from the "Detected objects" list above. Do not hallucinate additional objects.
 {get_language_prompt()}

web/src/components/overlay/ObjectTrackOverlay.tsx

@@ -12,38 +12,80 @@ import { TooltipPortal } from "@radix-ui/react-tooltip";
 import { cn } from "@/lib/utils";
 import { useTranslation } from "react-i18next";
 import { resolveZoneName } from "@/hooks/use-zone-friendly-name";
+import { Event } from "@/types/event";
 
 type ObjectTrackOverlayProps = {
   camera: string;
-  selectedObjectId: string;
   showBoundingBoxes?: boolean;
   currentTime: number;
   videoWidth: number;
   videoHeight: number;
   className?: string;
   onSeekToTime?: (timestamp: number, play?: boolean) => void;
-  objectTimeline?: ObjectLifecycleSequence[];
+};
+
+type PathPoint = {
+  x: number;
+  y: number;
+  timestamp: number;
+  lifecycle_item?: ObjectLifecycleSequence;
+  objectId: string;
+};
+
+type ObjectData = {
+  objectId: string;
+  label: string;
+  color: string;
+  pathPoints: PathPoint[];
+  currentZones: string[];
+  currentBox?: number[];
 };
 
 export default function ObjectTrackOverlay({
   camera,
-  selectedObjectId,
   showBoundingBoxes = false,
   currentTime,
   videoWidth,
   videoHeight,
   className,
   onSeekToTime,
-  objectTimeline,
 }: ObjectTrackOverlayProps) {
   const { t } = useTranslation("views/events");
   const { data: config } = useSWR<FrigateConfig>("config");
-  const { annotationOffset } = useDetailStream();
+  const { annotationOffset, selectedObjectIds } = useDetailStream();
 
   const effectiveCurrentTime = currentTime - annotationOffset / 1000;
 
-  // Fetch the full event data to get saved path points
-  const { data: eventData } = useSWR(["event_ids", { ids: selectedObjectId }]);
+  // Fetch all event data in a single request (CSV ids)
+  const { data: eventsData } = useSWR<Event[]>(
+    selectedObjectIds.length > 0
+      ? ["event_ids", { ids: selectedObjectIds.join(",") }]
+      : null,
+  );
+
+  // Fetch timeline data for each object ID using fixed number of hooks
+  const { data: timelineData } = useSWR<ObjectLifecycleSequence[]>(
+    selectedObjectIds.length > 0
+      ? `timeline?source_id=${selectedObjectIds.join(",")}&limit=1000`
+      : null,
+    { revalidateOnFocus: false },
+  );
+
+  const timelineResults = useMemo(() => {
+    // Group timeline entries by source_id
+    if (!timelineData) return selectedObjectIds.map(() => []);
+
+    const grouped: Record<string, ObjectLifecycleSequence[]> = {};
+    for (const entry of timelineData) {
+      if (!grouped[entry.source_id]) {
+        grouped[entry.source_id] = [];
+      }
+      grouped[entry.source_id].push(entry);
+    }
+
+    // Return timeline arrays in the same order as selectedObjectIds
+    return selectedObjectIds.map((id) => grouped[id] || []);
+  }, [selectedObjectIds, timelineData]);
 
   const typeColorMap = useMemo(
     () => ({
@@ -59,16 +101,18 @@ export default function ObjectTrackOverlay({
     [],
   );
 
-  const getObjectColor = useMemo(() => {
-    return (label: string) => {
+  const getObjectColor = useCallback(
+    (label: string, objectId: string) => {
       const objectColor = config?.model?.colormap[label];
       if (objectColor) {
         const reversed = [...objectColor].reverse();
         return `rgb(${reversed.join(",")})`;
       }
-      return "rgb(255, 0, 0)"; // fallback red
-    };
-  }, [config]);
+      // Fallback to deterministic color based on object ID
+      return generateColorFromId(objectId);
+    },
+    [config],
+  );
 
   const getZoneColor = useCallback(
     (zoneName: string) => {
@@ -82,128 +126,126 @@ export default function ObjectTrackOverlay({
     [config, camera],
   );
 
-  const currentObjectZones = useMemo(() => {
-    if (!objectTimeline) return [];
-
-    // Find the most recent timeline event at or before effective current time
-    const relevantEvents = objectTimeline
-      .filter((event) => event.timestamp <= effectiveCurrentTime)
-      .sort((a, b) => b.timestamp - a.timestamp); // Most recent first
-
-    // Get zones from the most recent event
-    return relevantEvents[0]?.data?.zones || [];
-  }, [objectTimeline, effectiveCurrentTime]);
-
-  const zones = useMemo(() => {
-    if (!config?.cameras?.[camera]?.zones || !currentObjectZones.length)
+  // Build per-object data structures
+  const objectsData = useMemo<ObjectData[]>(() => {
+    if (!eventsData || !Array.isArray(eventsData)) return [];
+    if (config?.cameras[camera]?.onvif.autotracking.enabled_in_config)
       return [];
 
+    return selectedObjectIds
+      .map((objectId, index) => {
+        const eventData = eventsData.find((e) => e.id === objectId);
+        const timelineData = timelineResults[index];
+
+        // get saved path points from event
+        const savedPathPoints: PathPoint[] =
+          eventData?.data?.path_data?.map(
+            ([coords, timestamp]: [number[], number]) => ({
+              x: coords[0],
+              y: coords[1],
+              timestamp,
+              lifecycle_item: undefined,
+              objectId,
+            }),
+          ) || [];
+
+        // timeline points for this object
+        const eventSequencePoints: PathPoint[] =
+          timelineData
+            ?.filter(
+              (event: ObjectLifecycleSequence) => event.data.box !== undefined,
+            )
+            .map((event: ObjectLifecycleSequence) => {
+              const [left, top, width, height] = event.data.box!;
+              event.data.zones_friendly_names = event?.data?.zones?.map(
+                (zone) => {
+                  return resolveZoneName(config, zone);
+                },
+              );
+              return {
+                x: left + width / 2, // Center x
+                y: top + height, // Bottom y
+                timestamp: event.timestamp,
+                lifecycle_item: event,
+                objectId,
+              };
+            }) || [];
+
+        // show full path once current time has reached the object's start time
+        const combinedPoints = [...savedPathPoints, ...eventSequencePoints]
+          .sort((a, b) => a.timestamp - b.timestamp)
+          .filter(
+            (point) =>
+              currentTime >= (eventData?.start_time ?? 0) &&
+              point.timestamp >= (eventData?.start_time ?? 0) &&
+              point.timestamp <= (eventData?.end_time ?? Infinity),
+          );
+
+        // Get color for this object
+        const label = eventData?.label || "unknown";
+        const color = getObjectColor(label, objectId);
+
+        // Get current zones
+        const currentZones =
+          timelineData
+            ?.filter(
+              (event: ObjectLifecycleSequence) =>
+                event.timestamp <= effectiveCurrentTime,
+            )
+            .sort(
+              (a: ObjectLifecycleSequence, b: ObjectLifecycleSequence) =>
+                b.timestamp - a.timestamp,
+            )[0]?.data?.zones || [];
+
+        // Get current bounding box
+        const currentBox = timelineData
+          ?.filter(
+            (event: ObjectLifecycleSequence) =>
+              event.timestamp <= effectiveCurrentTime && event.data.box,
+          )
+          .sort(
+            (a: ObjectLifecycleSequence, b: ObjectLifecycleSequence) =>
+              b.timestamp - a.timestamp,
+          )[0]?.data?.box;
+
+        return {
+          objectId,
+          label,
+          color,
+          pathPoints: combinedPoints,
+          currentZones,
+          currentBox,
+        };
+      })
+      .filter((obj: ObjectData) => obj.pathPoints.length > 0); // Only include objects with path data
+  }, [
+    eventsData,
+    selectedObjectIds,
+    timelineResults,
+    currentTime,
+    effectiveCurrentTime,
+    getObjectColor,
+    config,
+    camera,
+  ]);
+
+  // Collect all zones across all objects
+  const allZones = useMemo(() => {
+    if (!config?.cameras?.[camera]?.zones) return [];
+
+    const zoneNames = new Set<string>();
+    objectsData.forEach((obj) => {
+      obj.currentZones.forEach((zone) => zoneNames.add(zone));
+    });
+
     return Object.entries(config.cameras[camera].zones)
-      .filter(([name]) => currentObjectZones.includes(name))
+      .filter(([name]) => zoneNames.has(name))
       .map(([name, zone]) => ({
         name,
         coordinates: zone.coordinates,
         color: getZoneColor(name),
       }));
-  }, [config, camera, getZoneColor, currentObjectZones]);
-
-  // get saved path points from event
-  const savedPathPoints = useMemo(() => {
-    return (
-      eventData?.[0].data?.path_data?.map(
-        ([coords, timestamp]: [number[], number]) => ({
-          x: coords[0],
-          y: coords[1],
-          timestamp,
-          lifecycle_item: undefined,
-        }),
-      ) || []
-    );
-  }, [eventData]);
-
-  // timeline points for selected event
-  const eventSequencePoints = useMemo(() => {
-    return (
-      objectTimeline
-        ?.filter((event) => event.data.box !== undefined)
-        .map((event) => {
-          const [left, top, width, height] = event.data.box!;
-          event.data.zones_friendly_names = event?.data?.zones?.map((zone) => {
-            return resolveZoneName(config, zone);
-          });
-
-          return {
-            x: left + width / 2, // Center x
-            y: top + height, // Bottom y
-            timestamp: event.timestamp,
-            lifecycle_item: event,
-          };
-        }) || []
-    );
-  }, [config, objectTimeline]);
-
-  // final object path with timeline points included
-  const pathPoints = useMemo(() => {
-    // don't display a path for autotracking cameras
-    if (config?.cameras[camera]?.onvif.autotracking.enabled_in_config)
-      return [];
-
-    const combinedPoints = [...savedPathPoints, ...eventSequencePoints].sort(
-      (a, b) => a.timestamp - b.timestamp,
-    );
-
-    // Filter points around current time (within a reasonable window)
-    const timeWindow = 30; // 30 seconds window
-    return combinedPoints.filter(
-      (point) =>
-        point.timestamp >= currentTime - timeWindow &&
-        point.timestamp <= currentTime + timeWindow,
-    );
-  }, [savedPathPoints, eventSequencePoints, config, camera, currentTime]);
-
-  // get absolute positions on the svg canvas for each point
-  const absolutePositions = useMemo(() => {
-    if (!pathPoints) return [];
-
-    return pathPoints.map((point) => {
-      // Find the corresponding timeline entry for this point
-      const timelineEntry = objectTimeline?.find(
-        (entry) => entry.timestamp == point.timestamp,
-      );
-      return {
-        x: point.x * videoWidth,
-        y: point.y * videoHeight,
-        timestamp: point.timestamp,
-        lifecycle_item:
-          timelineEntry ||
-          (point.box // normal path point
-            ? {
-                timestamp: point.timestamp,
-                camera: camera,
-                source: "tracked_object",
-                source_id: selectedObjectId,
-                class_type: "visible" as LifecycleClassType,
-                data: {
-                  camera: camera,
-                  label: point.label,
-                  sub_label: "",
-                  box: point.box,
-                  region: [0, 0, 0, 0], // placeholder
-                  attribute: "",
-                  zones: [],
-                },
-              }
-            : undefined),
-      };
-    });
-  }, [
-    pathPoints,
-    videoWidth,
-    videoHeight,
-    objectTimeline,
-    camera,
-    selectedObjectId,
-  ]);
+  }, [config, camera, objectsData, getZoneColor]);
 
   const generateStraightPath = useCallback(
     (points: { x: number; y: number }[]) => {
@@ -218,15 +260,20 @@ export default function ObjectTrackOverlay({
   );
 
   const getPointColor = useCallback(
-    (baseColor: number[], type?: string) => {
+    (baseColorString: string, type?: string) => {
       if (type && typeColorMap[type as keyof typeof typeColorMap]) {
         const typeColor = typeColorMap[type as keyof typeof typeColorMap];
         if (typeColor) {
           return `rgb(${typeColor.join(",")})`;
         }
       }
-      // normal path point
-      return `rgb(${baseColor.map((c) => Math.max(0, c - 10)).join(",")})`;
+      // Parse and darken base color slightly for path points
+      const match = baseColorString.match(/\d+/g);
+      if (match) {
+        const [r, g, b] = match.map(Number);
+        return `rgb(${Math.max(0, r - 10)}, ${Math.max(0, g - 10)}, ${Math.max(0, b - 10)})`;
+      }
+      return baseColorString;
     },
     [typeColorMap],
   );
@@ -238,49 +285,8 @@ export default function ObjectTrackOverlay({
     [onSeekToTime],
   );
 
-  // render bounding box for object at current time if we have a timeline entry
-  const currentBoundingBox = useMemo(() => {
-    if (!objectTimeline) return null;
-
-    // Find the most recent timeline event at or before effective current time with a bounding box
-    const relevantEvents = objectTimeline
-      .filter(
-        (event) => event.timestamp <= effectiveCurrentTime && event.data.box,
-      )
-      .sort((a, b) => b.timestamp - a.timestamp); // Most recent first
-
-    const currentEvent = relevantEvents[0];
-
-    if (!currentEvent?.data.box) return null;
-
-    const [left, top, width, height] = currentEvent.data.box;
-    return {
-      left,
-      top,
-      width,
-      height,
-      centerX: left + width / 2,
-      centerY: top + height,
-    };
-  }, [objectTimeline, effectiveCurrentTime]);
-
-  const objectColor = useMemo(() => {
-    return pathPoints[0]?.label
-      ? getObjectColor(pathPoints[0].label)
-      : "rgb(255, 0, 0)";
-  }, [pathPoints, getObjectColor]);
-
-  const objectColorArray = useMemo(() => {
-    return pathPoints[0]?.label
-      ? getObjectColor(pathPoints[0].label).match(/\d+/g)?.map(Number) || [
-          255, 0, 0,
-        ]
-      : [255, 0, 0];
-  }, [pathPoints, getObjectColor]);
-
-  // render any zones for object at current time
   const zonePolygons = useMemo(() => {
-    return zones.map((zone) => {
+    return allZones.map((zone) => {
       // Convert zone coordinates from normalized (0-1) to pixel coordinates
       const points = zone.coordinates
         .split(",")
@@ -302,9 +308,9 @@ export default function ObjectTrackOverlay({
         stroke: zone.color,
       };
     });
-  }, [zones, videoWidth, videoHeight]);
+  }, [allZones, videoWidth, videoHeight]);
 
-  if (!pathPoints.length || !config) {
+  if (objectsData.length === 0 || !config) {
     return null;
   }
 
@@ -329,73 +335,102 @@ export default function ObjectTrackOverlay({
         />
       ))}
 
-      {absolutePositions.length > 1 && (
-        <path
-          d={generateStraightPath(absolutePositions)}
-          fill="none"
-          stroke={objectColor}
-          strokeWidth="5"
-          strokeLinecap="round"
-          strokeLinejoin="round"
-        />
-      )}
+      {objectsData.map((objData) => {
+        const absolutePositions = objData.pathPoints.map((point) => ({
+          x: point.x * videoWidth,
+          y: point.y * videoHeight,
+          timestamp: point.timestamp,
+          lifecycle_item: point.lifecycle_item,
+        }));
 
-      {absolutePositions.map((pos, index) => (
-        <Tooltip key={`point-${index}`}>
-          <TooltipTrigger asChild>
-            <circle
-              cx={pos.x}
-              cy={pos.y}
-              r="7"
-              fill={getPointColor(
-                objectColorArray,
-                pos.lifecycle_item?.class_type,
-              )}
-              stroke="white"
-              strokeWidth="3"
-              style={{ cursor: onSeekToTime ? "pointer" : "default" }}
-              onClick={() => handlePointClick(pos.timestamp)}
-            />
-          </TooltipTrigger>
-          <TooltipPortal>
-            <TooltipContent side="top" className="smart-capitalize">
-              {pos.lifecycle_item
-                ? `${pos.lifecycle_item.class_type.replace("_", " ")} at ${new Date(pos.timestamp * 1000).toLocaleTimeString()}`
-                : t("objectTrack.trackedPoint")}
-              {onSeekToTime && (
-                <div className="mt-1 text-xs text-muted-foreground">
-                  {t("objectTrack.clickToSeek")}
-                </div>
-              )}
-            </TooltipContent>
-          </TooltipPortal>
-        </Tooltip>
-      ))}
+        return (
+          <g key={objData.objectId}>
+            {absolutePositions.length > 1 && (
+              <path
+                d={generateStraightPath(absolutePositions)}
+                fill="none"
+                stroke={objData.color}
+                strokeWidth="5"
+                strokeLinecap="round"
+                strokeLinejoin="round"
+              />
+            )}
 
-      {currentBoundingBox && showBoundingBoxes && (
-        <g>
-          <rect
-            x={currentBoundingBox.left * videoWidth}
-            y={currentBoundingBox.top * videoHeight}
-            width={currentBoundingBox.width * videoWidth}
-            height={currentBoundingBox.height * videoHeight}
-            fill="none"
-            stroke={objectColor}
-            strokeWidth="5"
-            opacity="0.9"
-          />
+            {absolutePositions.map((pos, index) => (
+              <Tooltip key={`${objData.objectId}-point-${index}`}>
+                <TooltipTrigger asChild>
+                  <circle
+                    cx={pos.x}
+                    cy={pos.y}
+                    r="7"
+                    fill={getPointColor(
+                      objData.color,
+                      pos.lifecycle_item?.class_type,
+                    )}
+                    stroke="white"
+                    strokeWidth="3"
+                    style={{ cursor: onSeekToTime ? "pointer" : "default" }}
+                    onClick={() => handlePointClick(pos.timestamp)}
+                  />
+                </TooltipTrigger>
+                <TooltipPortal>
+                  <TooltipContent side="top" className="smart-capitalize">
+                    {pos.lifecycle_item
+                      ? `${pos.lifecycle_item.class_type.replace("_", " ")} at ${new Date(pos.timestamp * 1000).toLocaleTimeString()}`
+                      : t("objectTrack.trackedPoint")}
+                    {onSeekToTime && (
+                      <div className="mt-1 text-xs normal-case text-muted-foreground">
+                        {t("objectTrack.clickToSeek")}
+                      </div>
+                    )}
+                  </TooltipContent>
+                </TooltipPortal>
+              </Tooltip>
+            ))}
 
-          <circle
-            cx={currentBoundingBox.centerX * videoWidth}
-            cy={currentBoundingBox.centerY * videoHeight}
-            r="5"
-            fill="rgb(255, 255, 0)" // yellow highlight
-            stroke={objectColor}
-            strokeWidth="5"
-            opacity="1"
-          />
-        </g>
-      )}
+            {objData.currentBox && showBoundingBoxes && (
+              <g>
+                <rect
+                  x={objData.currentBox[0] * videoWidth}
+                  y={objData.currentBox[1] * videoHeight}
+                  width={objData.currentBox[2] * videoWidth}
+                  height={objData.currentBox[3] * videoHeight}
+                  fill="none"
+                  stroke={objData.color}
+                  strokeWidth="5"
+                  opacity="0.9"
+                />
+                <circle
+                  cx={
+                    (objData.currentBox[0] + objData.currentBox[2] / 2) *
+                    videoWidth
+                  }
+                  cy={
+                    (objData.currentBox[1] + objData.currentBox[3]) *
+                    videoHeight
+                  }
+                  r="5"
+                  fill="rgb(255, 255, 0)" // yellow highlight
+                  stroke={objData.color}
+                  strokeWidth="5"
+                  opacity="1"
+                />
+              </g>
+            )}
+          </g>
+        );
+      })}
     </svg>
   );
 }
+
+// Generate a deterministic HSL color from a string (object ID)
+function generateColorFromId(id: string): string {
+  let hash = 0;
+  for (let i = 0; i < id.length; i++) {
+    hash = id.charCodeAt(i) + ((hash << 5) - hash);
+  }
+  // Use golden ratio to distribute hues evenly
+  const hue = (hash * 137.508) % 360;
+  return `hsl(${hue}, 70%, 50%)`;
+}

web/src/components/overlay/detail/ObjectLifecycle.tsx

@@ -41,7 +41,7 @@ import {
   ContextMenuItem,
   ContextMenuTrigger,
 } from "@/components/ui/context-menu";
-import { useNavigate } from "react-router-dom";
+import { Link, useNavigate } from "react-router-dom";
 import { ObjectPath } from "./ObjectPath";
 import { getLifecycleItemDescription } from "@/utils/lifecycleUtil";
 import { IoPlayCircleOutline } from "react-icons/io5";
@@ -100,6 +100,10 @@ export default function ObjectLifecycle({
     );
   }, [config, event]);
 
+  const label = event.sub_label
+    ? event.sub_label
+    : getTranslatedLabel(event.label);
+
   const getZoneColor = useCallback(
     (zoneName: string) => {
       const zoneColor =
@@ -291,10 +295,10 @@ export default function ObjectLifecycle({
         timezone: config.ui.timezone,
         date_format:
           config.ui.time_format == "24hour"
-            ? t("time.formattedTimestampHourMinuteSecond.24hour", {
+            ? t("time.formattedTimestamp.24hour", {
                 ns: "common",
               })
-            : t("time.formattedTimestampHourMinuteSecond.12hour", {
+            : t("time.formattedTimestamp.12hour", {
                 ns: "common",
               }),
         time_style: "medium",
@@ -307,10 +311,10 @@ export default function ObjectLifecycle({
         timezone: config.ui.timezone,
         date_format:
           config.ui.time_format == "24hour"
-            ? t("time.formattedTimestampHourMinuteSecond.24hour", {
+            ? t("time.formattedTimestamp.24hour", {
                 ns: "common",
               })
-            : t("time.formattedTimestampHourMinuteSecond.12hour", {
+            : t("time.formattedTimestamp.12hour", {
                 ns: "common",
               }),
         time_style: "medium",
@@ -414,6 +418,7 @@ export default function ObjectLifecycle({
 
   return (
     <div className={className}>
+      <span tabIndex={0} className="sr-only" />
       {!fullscreen && (
         <div className={cn("flex items-center gap-2")}>
           <Button
@@ -634,17 +639,34 @@ export default function ObjectLifecycle({
               }}
               role="button"
             >
-              <div className={cn("ml-1 rounded-full bg-muted-foreground p-2")}>
+              <div
+                className={cn(
+                  "relative ml-2 rounded-full bg-muted-foreground p-2",
+                )}
+              >
                 {getIconForLabel(
-                  event.label,
-                  "size-6 text-primary dark:text-white",
+                  event.sub_label ? event.label + "-verified" : event.label,
+                  "size-4 text-white",
                 )}
               </div>
-              <div className="flex items-end gap-2">
-                <span>{getTranslatedLabel(event.label)}</span>
+              <div className="flex items-center gap-2">
+                <span className="capitalize">{label}</span>
                 <span className="text-secondary-foreground">
                   {formattedStart ?? ""} - {formattedEnd ?? ""}
                 </span>
+                {event.data?.recognized_license_plate && (
+                  <>
+                    <span className="text-secondary-foreground">·</span>
+                    <div className="text-sm text-secondary-foreground">
+                      <Link
+                        to={`/explore?recognized_license_plate=${event.data.recognized_license_plate}`}
+                        className="text-sm"
+                      >
+                        {event.data.recognized_license_plate}
+                      </Link>
+                    </div>
+                  </>
+                )}
               </div>
             </div>
           </div>
@@ -822,12 +844,12 @@ function LifecycleIconRow({
           />
         </div>
 
-        <div className="flex w-full flex-row justify-between">
+        <div className="ml-2 flex w-full min-w-0 flex-1">
           <div className="flex flex-col">
-            <Trans>
-              <div>{getLifecycleItemDescription(item)}</div>
-            </Trans>
-            <div className="mt-1 flex flex-wrap items-center gap-2 text-sm text-secondary-foreground md:gap-5">
+            <div className="text-md flex items-start break-words text-left">
+              <Trans>{getLifecycleItemDescription(item)}</Trans>
+            </div>
+            <div className="mt-1 flex flex-wrap items-center gap-2 text-xs text-secondary-foreground md:gap-5">
               <div className="flex items-center gap-1">
                 <span className="text-primary-variant">
                   {t("objectLifecycle.lifecycleItemDesc.header.ratio")}
@@ -886,8 +908,9 @@ function LifecycleIconRow({
               )}
             </div>
           </div>
-
-          <div className={cn("p-1 text-sm")}>{formattedEventTimestamp}</div>
+        </div>
+        <div className="ml-3 flex-shrink-0 px-1 text-right text-xs text-primary-variant">
+          <div className="whitespace-nowrap">{formattedEventTimestamp}</div>
         </div>
       </div>
     </div>

web/src/components/player/HlsVideoPlayer.tsx

@@ -20,7 +20,6 @@ import { cn } from "@/lib/utils";
 import { ASPECT_VERTICAL_LAYOUT, RecordingPlayerError } from "@/types/record";
 import { useTranslation } from "react-i18next";
 import ObjectTrackOverlay from "@/components/overlay/ObjectTrackOverlay";
-import { DetailStreamContextType } from "@/context/detail-stream-context";
 
 // Android native hls does not seek correctly
 const USE_NATIVE_HLS = !isAndroid;
@@ -54,8 +53,11 @@ type HlsVideoPlayerProps = {
   onUploadFrame?: (playTime: number) => Promise<AxiosResponse> | undefined;
   toggleFullscreen?: () => void;
   onError?: (error: RecordingPlayerError) => void;
-  detail?: Partial<DetailStreamContextType>;
+  isDetailMode?: boolean;
+  camera?: string;
+  currentTimeOverride?: number;
 };
+
 export default function HlsVideoPlayer({
   videoRef,
   containerRef,
@@ -75,17 +77,15 @@ export default function HlsVideoPlayer({
   onUploadFrame,
   toggleFullscreen,
   onError,
-  detail,
+  isDetailMode = false,
+  camera,
+  currentTimeOverride,
 }: HlsVideoPlayerProps) {
   const { t } = useTranslation("components/player");
   const { data: config } = useSWR<FrigateConfig>("config");
 
   // for detail stream context in History
-  const selectedObjectId = detail?.selectedObjectId;
-  const selectedObjectTimeline = detail?.selectedObjectTimeline;
-  const currentTime = detail?.currentTime;
-  const camera = detail?.camera;
-  const isDetailMode = detail?.isDetailMode ?? false;
+  const currentTime = currentTimeOverride;
 
   // playback
 
@@ -316,16 +316,14 @@ export default function HlsVideoPlayer({
         }}
       >
         {isDetailMode &&
-          selectedObjectId &&
           camera &&
           currentTime &&
           videoDimensions.width > 0 &&
           videoDimensions.height > 0 && (
             <div className="absolute z-50 size-full">
               <ObjectTrackOverlay
-                key={`${selectedObjectId}-${currentTime}`}
+                key={`overlay-${currentTime}`}
                 camera={camera}
-                selectedObjectId={selectedObjectId}
                 showBoundingBoxes={!isPlaying}
                 currentTime={currentTime}
                 videoWidth={videoDimensions.width}
@@ -336,7 +334,6 @@ export default function HlsVideoPlayer({
                     onSeekToTime(timestamp, play);
                   }
                 }}
-                objectTimeline={selectedObjectTimeline}
               />
             </div>
           )}

web/src/components/player/dynamic/DynamicVideoPlayer.tsx

@@ -61,7 +61,11 @@ export default function DynamicVideoPlayer({
   const { data: config } = useSWR<FrigateConfig>("config");
 
   // for detail stream context in History
-  const detail = useDetailStream();
+  const {
+    isDetailMode,
+    camera: contextCamera,
+    currentTime,
+  } = useDetailStream();
 
   // controlling playback
 
@@ -295,7 +299,9 @@ export default function DynamicVideoPlayer({
             setIsBuffering(true);
           }
         }}
-        detail={detail}
+        isDetailMode={isDetailMode}
+        camera={contextCamera || camera}
+        currentTimeOverride={currentTime}
       />
       <PreviewPlayer
         className={cn(

web/src/components/timeline/DetailStream.tsx

@@ -23,6 +23,7 @@ import { FrigatePlusDialog } from "@/components/overlay/dialog/FrigatePlusDialog
 import { cn } from "@/lib/utils";
 import { resolveZoneName } from "@/hooks/use-zone-friendly-name";
 import { Tooltip, TooltipContent, TooltipTrigger } from "../ui/tooltip";
+import { Link } from "react-router-dom";
 
 type DetailStreamProps = {
   reviewItems?: ReviewSegment[];
@@ -172,7 +173,11 @@ export default function DetailStream({
       <FrigatePlusDialog
         upload={upload}
         onClose={() => setUpload(undefined)}
-        onEventUploaded={() => setUpload(undefined)}
+        onEventUploaded={() => {
+          if (upload) {
+            upload.plus_id = "new_upload";
+          }
+        }}
       />
 
       <div
@@ -255,7 +260,9 @@ function ReviewGroup({
 
   const rawIconLabels: string[] = [
     ...(fetchedEvents
-      ? fetchedEvents.map((e) => e.label)
+      ? fetchedEvents.map((e) =>
+          e.sub_label ? e.label + "-verified" : e.label,
+        )
       : (review.data?.objects ?? [])),
     ...(review.data?.audio ?? []),
   ];
@@ -318,7 +325,7 @@ function ReviewGroup({
           <div className="ml-1 flex flex-col items-start gap-1.5">
             <div className="flex flex-row gap-3">
               <div className="text-sm font-medium">{displayTime}</div>
-              <div className="flex items-center gap-2">
+              <div className="relative flex items-center gap-2 text-white">
                 {iconLabels.slice(0, 5).map((lbl, idx) => (
                   <div
                     key={`${lbl}-${idx}`}
@@ -424,30 +431,34 @@ function EventList({
 }: EventListProps) {
   const { data: config } = useSWR<FrigateConfig>("config");
 
-  const { selectedObjectId, setSelectedObjectId } = useDetailStream();
+  const { selectedObjectIds, toggleObjectSelection } = useDetailStream();
+
+  const isSelected = selectedObjectIds.includes(event.id);
+
+  const label = event.sub_label || getTranslatedLabel(event.label);
 
   const handleObjectSelect = (event: Event | undefined) => {
     if (event) {
-      onSeek(event.start_time ?? 0);
-      setSelectedObjectId(event.id);
+      // onSeek(event.start_time ?? 0);
+      toggleObjectSelection(event.id);
     } else {
-      setSelectedObjectId(undefined);
+      toggleObjectSelection(undefined);
     }
   };
 
-  // Clear selectedObjectId when effectiveTime has passed this event's end_time
+  // Clear selection when effectiveTime has passed this event's end_time
   useEffect(() => {
-    if (selectedObjectId === event.id && effectiveTime && event.end_time) {
+    if (isSelected && effectiveTime && event.end_time) {
       if (effectiveTime >= event.end_time) {
-        setSelectedObjectId(undefined);
+        toggleObjectSelection(event.id);
       }
     }
   }, [
-    selectedObjectId,
+    isSelected,
     event.id,
     event.end_time,
     effectiveTime,
-    setSelectedObjectId,
+    toggleObjectSelection,
   ]);
 
   return (
@@ -455,48 +466,66 @@ function EventList({
       <div
         className={cn(
           "rounded-md bg-secondary p-2",
-          event.id == selectedObjectId
+          isSelected
             ? "bg-secondary-highlight"
             : "outline-transparent duration-500",
-          event.id != selectedObjectId &&
+          !isSelected &&
             (effectiveTime ?? 0) >= (event.start_time ?? 0) - 0.5 &&
             (effectiveTime ?? 0) <=
               (event.end_time ?? event.start_time ?? 0) + 0.5 &&
             "bg-secondary-highlight",
         )}
       >
-        <div className="ml-1.5 flex w-full items-center justify-between">
-          <div
-            className="flex items-center gap-2 text-sm font-medium"
-            onClick={(e) => {
-              e.stopPropagation();
-              handleObjectSelect(
-                event.id == selectedObjectId ? undefined : event,
-              );
-            }}
-            role="button"
-          >
+        <div className="ml-1.5 flex w-full items-end justify-between">
+          <div className="flex flex-1 items-center gap-2 text-sm font-medium">
             <div
               className={cn(
-                "rounded-full p-1",
-                event.id == selectedObjectId
-                  ? "bg-selected"
-                  : "bg-muted-foreground",
+                "relative rounded-full p-1 text-white",
+                isSelected ? "bg-selected" : "bg-muted-foreground",
               )}
+              onClick={(e) => {
+                e.stopPropagation();
+                handleObjectSelect(isSelected ? undefined : event);
+              }}
             >
-              {getIconForLabel(event.label, "size-3 text-white")}
+              {getIconForLabel(
+                event.sub_label ? event.label + "-verified" : event.label,
+                "size-3 text-white",
+              )}
             </div>
-            <div className="flex items-end gap-2">
-              <span>{getTranslatedLabel(event.label)}</span>
+            <div
+              className="flex flex-1 items-center gap-2"
+              onClick={(e) => {
+                e.stopPropagation();
+                onSeek(event.start_time ?? 0);
+              }}
+              role="button"
+            >
+              <div className="flex gap-2">
+                <span className="capitalize">{label}</span>
+                {event.data?.recognized_license_plate && (
+                  <>
+                    <span className="text-secondary-foreground">·</span>
+                    <div className="text-sm text-secondary-foreground">
+                      <Link
+                        to={`/explore?recognized_license_plate=${event.data.recognized_license_plate}`}
+                        className="text-sm"
+                      >
+                        {event.data.recognized_license_plate}
+                      </Link>
+                    </div>
+                  </>
+                )}
+              </div>
             </div>
           </div>
-          <div className="mr-2 flex flex-1 flex-row justify-end">
+          <div className="mr-2 flex flex-row justify-end">
             <EventMenu
               event={event}
               config={config}
               onOpenUpload={(e) => onOpenUpload?.(e)}
-              selectedObjectId={selectedObjectId}
-              setSelectedObjectId={handleObjectSelect}
+              isSelected={isSelected}
+              onToggleSelection={handleObjectSelect}
             />
           </div>
         </div>
@@ -599,10 +628,11 @@ function LifecycleItem({
           )}
         />
       </div>
-      <div className="flex w-full flex-row justify-between">
+
+      <div className="ml-0.5 flex min-w-0 flex-1">
         <Tooltip>
           <TooltipTrigger>
-            <div className="flex items-start text-left">
+            <div className="flex items-start break-words text-left">
               <Trans>{getLifecycleItemDescription(item)}</Trans>
             </div>
           </TooltipTrigger>
@@ -622,7 +652,9 @@ function LifecycleItem({
                   </span>
                   {areaPx !== undefined && areaPct !== undefined ? (
                     <span className="font-medium text-foreground">
-                      {areaPx} {t("pixels", { ns: "common" })} · {areaPct}%
+                      {areaPx} {t("pixels", { ns: "common" })}{" "}
+                      <span className="text-secondary-foreground">·</span>{" "}
+                      {areaPct}%
                     </span>
                   ) : (
                     <span>N/A</span>
@@ -632,7 +664,10 @@ function LifecycleItem({
             </div>
           </TooltipContent>
         </Tooltip>
-        <div className={cn("p-1 text-xs")}>{formattedEventTimestamp}</div>
+      </div>
+
+      <div className="ml-3 flex-shrink-0 px-1 text-right text-xs text-primary-variant">
+        <div className="whitespace-nowrap">{formattedEventTimestamp}</div>
       </div>
     </div>
   );

web/src/components/timeline/EventMenu.tsx

@@ -12,14 +12,15 @@ import { useNavigate } from "react-router-dom";
 import { useTranslation } from "react-i18next";
 import { Event } from "@/types/event";
 import { FrigateConfig } from "@/types/frigateConfig";
+import { useState } from "react";
 
 type EventMenuProps = {
   event: Event;
   config?: FrigateConfig;
   onOpenUpload?: (e: Event) => void;
   onOpenSimilarity?: (e: Event) => void;
-  selectedObjectId?: string;
-  setSelectedObjectId?: (event: Event | undefined) => void;
+  isSelected?: boolean;
+  onToggleSelection?: (event: Event | undefined) => void;
 };
 
 export default function EventMenu({
@@ -27,25 +28,26 @@ export default function EventMenu({
   config,
   onOpenUpload,
   onOpenSimilarity,
-  selectedObjectId,
-  setSelectedObjectId,
+  isSelected = false,
+  onToggleSelection,
 }: EventMenuProps) {
   const apiHost = useApiHost();
   const navigate = useNavigate();
   const { t } = useTranslation("views/explore");
+  const [isOpen, setIsOpen] = useState(false);
 
   const handleObjectSelect = () => {
-    if (event.id === selectedObjectId) {
-      setSelectedObjectId?.(undefined);
+    if (isSelected) {
+      onToggleSelection?.(undefined);
     } else {
-      setSelectedObjectId?.(event);
+      onToggleSelection?.(event);
     }
   };
 
   return (
     <>
       <span tabIndex={0} className="sr-only" />
-      <DropdownMenu>
+      <DropdownMenu open={isOpen} onOpenChange={setIsOpen}>
         <DropdownMenuTrigger>
           <div className="rounded p-1 pr-2" role="button">
             <HiDotsHorizontal className="size-4 text-muted-foreground" />
@@ -54,7 +56,7 @@ export default function EventMenu({
         <DropdownMenuPortal>
           <DropdownMenuContent>
             <DropdownMenuItem onSelect={handleObjectSelect}>
-              {event.id === selectedObjectId
+              {isSelected
                 ? t("itemMenu.hideObjectDetails.label")
                 : t("itemMenu.showObjectDetails.label")}
             </DropdownMenuItem>
@@ -85,6 +87,7 @@ export default function EventMenu({
               config?.plus?.enabled && (
                 <DropdownMenuItem
                   onSelect={() => {
+                    setIsOpen(false);
                     onOpenUpload?.(event);
                   }}
                 >

web/src/context/detail-stream-context.tsx

@@ -1,16 +1,14 @@
 import React, { createContext, useContext, useState, useEffect } from "react";
 import { FrigateConfig } from "@/types/frigateConfig";
 import useSWR from "swr";
-import { ObjectLifecycleSequence } from "@/types/timeline";
 
 export interface DetailStreamContextType {
-  selectedObjectId: string | undefined;
-  selectedObjectTimeline?: ObjectLifecycleSequence[];
+  selectedObjectIds: string[];
   currentTime: number;
   camera: string;
   annotationOffset: number; // milliseconds
   setAnnotationOffset: (ms: number) => void;
-  setSelectedObjectId: (id: string | undefined) => void;
+  toggleObjectSelection: (id: string | undefined) => void;
   isDetailMode: boolean;
 }
 
@@ -31,13 +29,21 @@ export function DetailStreamProvider({
   currentTime,
   camera,
 }: DetailStreamProviderProps) {
-  const [selectedObjectId, setSelectedObjectId] = useState<
-    string | undefined
-  >();
+  const [selectedObjectIds, setSelectedObjectIds] = useState<string[]>([]);
 
-  const { data: selectedObjectTimeline } = useSWR<ObjectLifecycleSequence[]>(
-    selectedObjectId ? ["timeline", { source_id: selectedObjectId }] : null,
-  );
+  const toggleObjectSelection = (id: string | undefined) => {
+    if (id === undefined) {
+      setSelectedObjectIds([]);
+    } else {
+      setSelectedObjectIds((prev) => {
+        if (prev.includes(id)) {
+          return prev.filter((existingId) => existingId !== id);
+        } else {
+          return [...prev, id];
+        }
+      });
+    }
+  };
 
   const { data: config } = useSWR<FrigateConfig>("config");
 
@@ -52,14 +58,18 @@ export function DetailStreamProvider({
     setAnnotationOffset(cfgOffset);
   }, [config, camera]);
 
+  // Clear selected objects when exiting detail mode or changing cameras
+  useEffect(() => {
+    setSelectedObjectIds([]);
+  }, [isDetailMode, camera]);
+
   const value: DetailStreamContextType = {
-    selectedObjectId,
-    selectedObjectTimeline,
+    selectedObjectIds,
     currentTime,
     camera,
     annotationOffset,
     setAnnotationOffset,
-    setSelectedObjectId,
+    toggleObjectSelection,
     isDetailMode,
   };
 

web/src/types/event.ts

@@ -22,6 +22,7 @@ export interface Event {
     area: number;
     ratio: number;
     type: "object" | "audio" | "manual";
+    recognized_license_plate?: string;
     path_data: [number[], number][];
   };
 }

web/src/utils/lifecycleUtil.ts

@@ -1,6 +1,7 @@
 import { ObjectLifecycleSequence } from "@/types/timeline";
 import { t } from "i18next";
 import i18n, { getTranslatedLabel } from "./i18n";
+import { capitalizeFirstLetter } from "./stringUtil";
 
 export function getLifecycleItemDescription(
   lifecycleItem: ObjectLifecycleSequence,
@@ -10,7 +11,7 @@ export function getLifecycleItemDescription(
     : lifecycleItem.data.sub_label || lifecycleItem.data.label;
 
   const label = lifecycleItem.data.sub_label
-    ? rawLabel
+    ? capitalizeFirstLetter(rawLabel)
     : getTranslatedLabel(rawLabel);
 
   let supportedListFormat = false;

web/src/views/recording/RecordingView.tsx

@@ -11,6 +11,7 @@ import DetailStream from "@/components/timeline/DetailStream";
 import { Button } from "@/components/ui/button";
 import { ToggleGroup, ToggleGroupItem } from "@/components/ui/toggle-group";
 import { useOverlayState } from "@/hooks/use-overlay-state";
+import { useResizeObserver } from "@/hooks/resize-observer";
 import { ExportMode } from "@/types/filter";
 import { FrigateConfig } from "@/types/frigateConfig";
 import { Preview } from "@/types/preview";
@@ -31,12 +32,7 @@ import {
   useRef,
   useState,
 } from "react";
-import {
-  isDesktop,
-  isMobile,
-  isMobileOnly,
-  isTablet,
-} from "react-device-detect";
+import { isDesktop, isMobile } from "react-device-detect";
 import { IoMdArrowRoundBack } from "react-icons/io";
 import { useNavigate } from "react-router-dom";
 import { Toaster } from "@/components/ui/sonner";
@@ -55,7 +51,6 @@ import {
   RecordingSegment,
   RecordingStartingPoint,
 } from "@/types/record";
-import { useResizeObserver } from "@/hooks/resize-observer";
 import { cn } from "@/lib/utils";
 import { useFullscreen } from "@/hooks/use-fullscreen";
 import { useTimezone } from "@/hooks/use-date-utils";
@@ -399,49 +394,47 @@ export function RecordingView({
     }
   }, [mainCameraAspect]);
 
-  const [{ width: mainWidth, height: mainHeight }] =
+  // use a resize observer to determine whether to use w-full or h-full based on container aspect ratio
+  const [{ width: containerWidth, height: containerHeight }] =
     useResizeObserver(cameraLayoutRef);
+  const [{ width: previewRowWidth, height: previewRowHeight }] =
+    useResizeObserver(previewRowRef);
 
-  const mainCameraStyle = useMemo(() => {
-    if (isMobile || mainCameraAspect != "normal" || !config) {
-      return undefined;
+  const useHeightBased = useMemo(() => {
+    if (!containerWidth || !containerHeight) {
+      return false;
     }
 
-    const camera = config.cameras[mainCamera];
-
-    if (!camera) {
-      return undefined;
+    const cameraAspectRatio = getCameraAspect(mainCamera);
+    if (!cameraAspectRatio) {
+      return false;
     }
 
-    const aspect = getCameraAspect(mainCamera);
+    // Calculate available space for camera after accounting for preview row
+    // For tall cameras: preview row is side-by-side (takes width)
+    // For wide/normal cameras: preview row is stacked (takes height)
+    const availableWidth =
+      mainCameraAspect == "tall" && previewRowWidth
+        ? containerWidth - previewRowWidth
+        : containerWidth;
+    const availableHeight =
+      mainCameraAspect != "tall" && previewRowHeight
+        ? containerHeight - previewRowHeight
+        : containerHeight;
 
-    if (!aspect) {
-      return undefined;
-    }
+    const availableAspectRatio = availableWidth / availableHeight;
 
-    const availableHeight = mainHeight - 112;
-
-    let percent;
-    if (mainWidth / availableHeight < aspect) {
-      percent = 100;
-    } else {
-      const availableWidth = aspect * availableHeight;
-      percent =
-        (mainWidth < availableWidth
-          ? mainWidth / availableWidth
-          : availableWidth / mainWidth) * 100;
-    }
-
-    return {
-      width: `${Math.round(percent)}%`,
-    };
+    // If available space is wider than camera aspect, constrain by height (h-full)
+    // If available space is taller than camera aspect, constrain by width (w-full)
+    return availableAspectRatio >= cameraAspectRatio;
   }, [
-    config,
-    mainCameraAspect,
-    mainWidth,
-    mainHeight,
-    mainCamera,
+    containerWidth,
+    containerHeight,
+    previewRowWidth,
+    previewRowHeight,
     getCameraAspect,
+    mainCamera,
+    mainCameraAspect,
   ]);
 
   const previewRowOverflows = useMemo(() => {
@@ -685,19 +678,17 @@ export function RecordingView({
         <div
           ref={mainLayoutRef}
           className={cn(
-            "flex h-full justify-center overflow-hidden",
-            isDesktop ? "" : "flex-col gap-2 landscape:flex-row",
+            "flex flex-1 overflow-hidden",
+            isDesktop ? "flex-row" : "flex-col gap-2 landscape:flex-row",
           )}
         >
           <div
             ref={cameraLayoutRef}
             className={cn(
-              "flex flex-1 flex-wrap",
+              "flex flex-1 flex-wrap overflow-hidden",
               isDesktop
-                ? timelineType === "detail"
-                  ? "md:w-[40%] lg:w-[70%] xl:w-full"
-                  : "w-[80%]"
-                : "",
+                ? "min-w-0 px-4"
+                : "portrait:max-h-[50dvh] portrait:flex-shrink-0 portrait:flex-grow-0 portrait:basis-auto",
             )}
           >
             <div
@@ -711,37 +702,24 @@ export function RecordingView({
               <div
                 key={mainCamera}
                 className={cn(
-                  "relative",
+                  "relative flex max-h-full min-h-0 min-w-0 max-w-full items-center justify-center",
                   isDesktop
-                    ? cn(
-                        "flex justify-center px-4",
-                        mainCameraAspect == "tall"
-                          ? "h-[50%] md:h-[60%] lg:h-[75%] xl:h-[90%]"
-                          : mainCameraAspect == "wide"
-                            ? "w-full"
-                            : "",
-                      )
+                    ? // Desktop: dynamically switch between w-full and h-full based on
+                      // container vs camera aspect ratio to ensure proper fitting
+                      useHeightBased
+                      ? "h-full"
+                      : "w-full"
                     : cn(
-                        "pt-2 portrait:w-full",
-                        isMobileOnly &&
-                          (mainCameraAspect == "wide"
-                            ? "aspect-wide landscape:w-full"
-                            : "aspect-video landscape:h-[94%] landscape:xl:h-[65%]"),
-                        isTablet &&
-                          (mainCameraAspect == "wide"
-                            ? "aspect-wide landscape:w-full"
-                            : mainCameraAspect == "normal"
-                              ? "landscape:w-full"
-                              : "aspect-video landscape:h-[100%]"),
+                        "flex-shrink-0 portrait:w-full landscape:h-full",
+                        mainCameraAspect == "wide"
+                          ? "aspect-wide"
+                          : mainCameraAspect == "tall"
+                            ? "aspect-tall portrait:h-full"
+                            : "aspect-video",
                       ),
                 )}
                 style={{
-                  width: mainCameraStyle ? mainCameraStyle.width : undefined,
-                  aspectRatio: isDesktop
-                    ? mainCameraAspect == "tall"
-                      ? getCameraAspect(mainCamera)
-                      : undefined
-                    : Math.max(1, getCameraAspect(mainCamera) ?? 0),
+                  aspectRatio: getCameraAspect(mainCamera),
                 }}
               >
                 {isDesktop && (
@@ -782,10 +760,10 @@ export function RecordingView({
                 <div
                   ref={previewRowRef}
                   className={cn(
-                    "scrollbar-container flex gap-2 overflow-auto",
+                    "scrollbar-container flex flex-shrink-0 gap-2 overflow-auto",
                     mainCameraAspect == "tall"
-                      ? "h-full w-72 flex-col"
-                      : `h-28 w-full`,
+                      ? "ml-2 h-full w-72 min-w-72 flex-col"
+                      : "h-28 min-h-28 w-full",
                     previewRowOverflows ? "" : "items-center justify-center",
                     timelineType == "detail" && isDesktop && "mt-4",
                   )}
@@ -971,10 +949,23 @@ function Timeline({
   return (
     <div
       className={cn(
-        "relative",
+        "relative overflow-hidden",
         isDesktop
-          ? `${timelineType == "timeline" ? "w-[100px]" : timelineType == "detail" ? "w-[30%] min-w-[350px]" : "w-60"} no-scrollbar overflow-y-auto`
-          : `overflow-hidden portrait:flex-grow ${timelineType == "timeline" ? "landscape:w-[100px]" : timelineType == "detail" && isDesktop ? "flex-1" : "landscape:w-[300px]"} `,
+          ? cn(
+              "no-scrollbar overflow-y-auto",
+              timelineType == "timeline"
+                ? "w-[100px] flex-shrink-0"
+                : timelineType == "detail"
+                  ? "min-w-[20rem] max-w-[30%] flex-shrink-0 flex-grow-0 basis-[30rem] md:min-w-[20rem] md:max-w-[25%] lg:min-w-[30rem] lg:max-w-[33%]"
+                  : "w-60 flex-shrink-0",
+            )
+          : cn(
+              timelineType == "timeline"
+                ? "portrait:flex-grow landscape:w-[100px] landscape:flex-shrink-0"
+                : timelineType == "detail"
+                  ? "portrait:flex-grow landscape:w-[19rem] landscape:flex-shrink-0"
+                  : "portrait:flex-grow landscape:w-[19rem] landscape:flex-shrink-0",
+            ),
       )}
     >
       {isMobile && (