change icon, make transition animation during analyzing image

src/components/playground/PhoneSimulator.tsx

@@ -28,17 +28,76 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
   const tracks = useTracks();
   const voiceAssistant = useVoiceAssistant();
   const fileInputRef = useRef<HTMLInputElement>(null);
+  const phoneContainerRef = useRef<HTMLDivElement>(null);
+  const visualizerRef = useRef<HTMLDivElement>(null);
   const [showCameraMenu, setShowCameraMenu] = useState(false);
   const [cameras, setCameras] = useState<MediaDeviceInfo[]>([]);
+  const [processingImage, setProcessingImage] = useState<string | null>(null);
+  const [isCapturing, setIsCapturing] = useState(false);
+  const [processingSource, setProcessingSource] = useState<
+    "camera" | "upload" | null
+  >(null);
 
   const [currentTime, setCurrentTime] = useState("");
 
+  const [visualizerPosition, setVisualizerPosition] = useState({
+    x: 16,
+    y: 56,
+  });
+  const [isDragging, setIsDragging] = useState(false);
+  const dragOffset = useRef({ x: 0, y: 0 });
+
+  const handleDragStart = (e: React.MouseEvent) => {
+    setIsDragging(true);
+    dragOffset.current = {
+      x: e.clientX - visualizerPosition.x,
+      y: e.clientY - visualizerPosition.y,
+    };
+  };
+
+  const handleDragMove = (e: MouseEvent) => {
+    if (!isDragging || !phoneContainerRef.current || !visualizerRef.current) return;
+
+    const containerRect = phoneContainerRef.current.getBoundingClientRect();
+    const visualizerRect = visualizerRef.current.getBoundingClientRect();
+
+    let newX = e.clientX - dragOffset.current.x;
+    let newY = e.clientY - dragOffset.current.y;
+
+    // Constrain within container
+    const maxX = containerRect.width - visualizerRect.width;
+    const maxY = containerRect.height - visualizerRect.height;
+    const statusBarHeight = 48; // h-12 = 48px
+
+    newX = Math.max(0, Math.min(newX, maxX));
+    newY = Math.max(statusBarHeight, Math.min(newY, maxY));
+
+    setVisualizerPosition({
+      x: newX,
+      y: newY,
+    });
+  };
+
+  const handleDragEnd = () => {
+    setIsDragging(false);
+  };
+
+  useEffect(() => {
+    if (isDragging) {
+      window.addEventListener("mouseup", handleDragEnd);
+      window.addEventListener("mousemove", handleDragMove);
+    }
+    return () => {
+      window.removeEventListener("mouseup", handleDragEnd);
+      window.removeEventListener("mousemove", handleDragMove);
+    };
+  }, [isDragging]);
+
   useEffect(() => {
     if (showCameraMenu) {
       Room.getLocalDevices("videoinput").then(setCameras);
     }
   }, [showCameraMenu]);
-
   // Close menu when clicking outside
   useEffect(() => {
     const handleClickOutside = (event: MouseEvent) => {
@@ -54,6 +113,13 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
     };
   }, [showCameraMenu]);
 
+  useEffect(() => {
+    if (voiceAssistant.state === "speaking") {
+      setProcessingImage(null);
+      setProcessingSource(null);
+    }
+  }, [voiceAssistant.state]);
+
   useEffect(() => {
     const updateTime = () => {
       const now = new Date();
@@ -92,22 +158,128 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
   };
 
   const handleCapture = async () => {
-    if (!localCameraTrack || !onCapture) return;
+    if (!localCameraTrack || !onCapture || isCapturing) return;
+    setIsCapturing(true);
+
+    const trackReference = localCameraTrack as any;
+    // Handle both TrackReference (from useTracks) and potential direct Track objects
+    const track =
+      trackReference.publication?.track?.mediaStreamTrack ||
+      trackReference.mediaStreamTrack;
+
+    if (!track) {
+      console.error("No media stream track found");
+      setIsCapturing(false);
+      return;
+    }
+
+    const video = document.createElement("video");
+    video.srcObject = new MediaStream([track]);
+    video.muted = true;
+    video.playsInline = true;
+    video.autoplay = true;
+    // Element needs to be in the DOM for some browsers to play it properly
+    video.style.position = "absolute";
+    video.style.top = "-9999px";
+    video.style.left = "-9999px";
+    document.body.appendChild(video);
+
+    try {
+      await video.play();
+
+      // Wait for video dimensions to be available
+      if (video.videoWidth === 0 || video.videoHeight === 0) {
+        await new Promise<void>((resolve) => {
+          video.onloadedmetadata = () => resolve();
+          // Timeout to prevent hanging
+          setTimeout(resolve, 1000);
+        });
+      }
 
-    const videoElement = localCameraTrack.attach() as HTMLVideoElement;
       const canvas = document.createElement("canvas");
-    canvas.width = videoElement.videoWidth;
-    canvas.height = videoElement.videoHeight;
+      // Default to video dimensions
+      let renderWidth = video.videoWidth;
+      let renderHeight = video.videoHeight;
+      let sourceX = 0;
+      let sourceY = 0;
+      let sourceWidth = video.videoWidth;
+      let sourceHeight = video.videoHeight;
+
+      // If the video is landscape but we want a portrait crop (like a phone)
+      // We want an aspect ratio of roughly 9:19.5 (from the container styles)
+      const targetAspect = 9 / 19.5;
+      const videoAspect = video.videoWidth / video.videoHeight;
+
+      if (videoAspect > targetAspect) {
+        // Video is wider than target - crop width
+        const newWidth = video.videoHeight * targetAspect;
+        sourceX = (video.videoWidth - newWidth) / 2;
+        sourceWidth = newWidth;
+        renderWidth = newWidth;
+      } else {
+        // Video is taller than target - crop height (less common for landscape webcam)
+        const newHeight = video.videoWidth / targetAspect;
+        sourceY = (video.videoHeight - newHeight) / 2;
+        sourceHeight = newHeight;
+        renderHeight = newHeight;
+      }
+
+      canvas.width = renderWidth;
+      canvas.height = renderHeight;
       const ctx = canvas.getContext("2d");
 
       if (ctx) {
-        ctx.drawImage(videoElement, 0, 0);
+        // Mirror the image to match the preview
+        ctx.translate(canvas.width, 0);
+        ctx.scale(-1, 1);
+
+        // Draw only the cropped portion of the video
+        ctx.drawImage(
+          video,
+          sourceX,
+          sourceY,
+          sourceWidth,
+          sourceHeight,
+          0,
+          0,
+          renderWidth,
+          renderHeight
+        );
+        // Reset transform
+        ctx.setTransform(1, 0, 0, 1, 0, 0);
+
+        // Use toDataURL for immediate preview feedback
+        const dataUrl = canvas.toDataURL("image/jpeg");
+        setProcessingImage(dataUrl);
+        setProcessingSource("camera");
+
+        // Create a new canvas for the final output (unmirrored if needed, but user requested mirrored)
+        // The user requested to mirror the shuttled photo, which we did above for the canvas.
+        // So the blob created from this canvas will also be mirrored.
+
         canvas.toBlob((blob) => {
           if (blob && onCapture) {
-                const file = new File([blob], "camera-capture.jpg", { type: "image/jpeg" });
+            const file = new File([blob], "camera-capture.jpg", {
+              type: "image/jpeg",
+            });
             onCapture(file);
           }
+          setIsCapturing(false);
         }, "image/jpeg");
+      } else {
+        setIsCapturing(false);
+      }
+    } catch (e) {
+      console.error("Failed to capture image", e);
+      setIsCapturing(false);
+    } finally {
+      // Cleanup
+      video.pause();
+      video.srcObject = null;
+      if (document.body.contains(video)) {
+        document.body.removeChild(video);
+      }
+      video.remove();
     }
   };
 
@@ -131,6 +303,8 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
     const file = event.target.files?.[0];
     if (file && onCapture) {
       onCapture(file);
+      setProcessingImage(URL.createObjectURL(file));
+      setProcessingSource("upload");
     }
     // Reset input so the same file can be selected again
     if (event.target) {
@@ -182,6 +356,14 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
         .mirror-video video {
             transform: scaleX(-1);
         }
+        @keyframes scan {
+          0% { top: 0%; }
+          50% { top: 100%; }
+          100% { top: 0%; }
+        }
+        .scan-animation {
+          animation: scan 3s linear infinite;
+        }
       `}</style>
       {/* Status Bar */}
       <div className="h-12 w-full bg-black/20 absolute top-0 left-0 z-20 flex items-center justify-between px-6 text-white text-xs font-medium backdrop-blur-sm">
@@ -196,8 +378,28 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
       <div className="absolute top-0 left-1/2 -translate-x-1/2 w-32 h-7 bg-black rounded-b-2xl z-30"></div>
 
       {/* Main Content */}
-      <div className="flex-grow relative bg-gray-950 w-full h-full">
+      <div ref={phoneContainerRef} className="flex-grow relative bg-gray-950 w-full h-full">
         {videoContent}
+        {processingImage && (
+          <div className="absolute inset-0 z-10 bg-black flex items-center justify-center">
+            {/* eslint-disable-next-line @next/next/no-img-element */}
+            <img
+              src={processingImage}
+              alt="Processing"
+              className={`w-full h-full opacity-50 ${
+                processingSource === "camera"
+                  ? "object-cover scale-x-[-1]"
+                  : "object-contain"
+              }`}
+            />
+            <div className="absolute inset-0 overflow-hidden">
+               <div className="w-full h-[2px] bg-blue-500 shadow-[0_0_15px_rgba(59,130,246,1)] scan-animation absolute"></div>
+            </div>
+            <div className="absolute top-1/2 left-1/2 -translate-x-1/2 -translate-y-1/2 w-full text-center text-blue-400 font-mono text-sm animate-pulse z-20">
+              正在分析...
+            </div>
+          </div>
+        )}
         
         <input 
             type="file" 
@@ -208,7 +410,7 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
         />
 
         {/* Capture Guide Lines */}
-        {roomState === ConnectionState.Connected && phoneMode === "capture" && (
+        {roomState === ConnectionState.Connected && phoneMode === "capture" && !processingImage && (
             <div className="absolute inset-0 pointer-events-none z-10">
                 {/* Thirds Grid */}
                 <div className="absolute top-1/3 left-0 w-full h-[1px] bg-white/20"></div>
@@ -222,9 +424,17 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
             </div>
         )}
         
-        {/* Agent Audio Visualizer (Top Left) */}
+        {/* Agent Audio Visualizer (Draggable) */}
         {roomState === ConnectionState.Connected && voiceAssistant.audioTrack && (
-            <div className="absolute top-14 left-4 z-20 p-2 bg-black/40 backdrop-blur-md rounded-lg border border-white/10 shadow-lg">
+          <div
+            ref={visualizerRef}
+            className="absolute z-20 p-2 bg-black/40 backdrop-blur-md rounded-lg border border-white/10 shadow-lg cursor-move select-none"
+            style={{
+              left: visualizerPosition.x,
+              top: visualizerPosition.y,
+            }}
+            onMouseDown={handleDragStart}
+          >
             <div className="h-8 w-24 flex items-center justify-center [--lk-va-bar-width:3px] [--lk-va-bar-gap:2px] [--lk-fg:white]">
               <BarVisualizer
                 state={voiceAssistant.state}
@@ -239,9 +449,9 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
 
       {/* Call Controls Overlay */}
       {roomState === ConnectionState.Connected && (
-        <div className="absolute bottom-8 left-0 w-full px-8 flex items-center justify-between z-20">
+        <div className="absolute bottom-8 left-0 w-full px-8 z-20">
           {phoneMode === "capture" ? (
-                <div className="w-full grid grid-cols-3 items-center px-8">
+            <div className="w-full grid grid-cols-3 items-center">
               <div className="flex justify-start">
                 <button
                   className="p-4 rounded-full bg-gray-800/50 text-white hover:bg-gray-800/70 transition-colors shrink-0"
@@ -268,7 +478,9 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
                 {showCameraMenu && (
                   <div className="absolute bottom-16 right-0 bg-gray-900 border border-gray-800 rounded-lg shadow-xl py-2 w-48 z-50">
                     {cameras.length === 0 ? (
-                                    <div className="px-4 py-2 text-gray-500 text-sm">No cameras found</div>
+                      <div className="px-4 py-2 text-gray-500 text-sm">
+                        No cameras found
+                      </div>
                     ) : (
                       cameras.map((device) => (
                         <button
@@ -276,7 +488,8 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
                           onClick={() => handleSelectCamera(device.deviceId)}
                           className="w-full text-left px-4 py-2 text-sm text-white hover:bg-gray-800 transition-colors truncate"
                         >
-                                            {device.label || `Camera ${cameras.indexOf(device) + 1}`}
+                          {device.label ||
+                            `Camera ${cameras.indexOf(device) + 1}`}
                         </button>
                       ))
                     )}
@@ -285,7 +498,7 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
               </div>
             </div>
           ) : (
-                <>
+            <div className="w-full flex items-center justify-center gap-8">
               <button
                 className={`p-4 rounded-full backdrop-blur-md transition-colors ${
                   !isMicEnabled
@@ -307,7 +520,7 @@ export function PhoneSimulator({ onConnect, phoneMode = "normal", onCapture }: P
               >
                 <PhoneOffIcon className="w-6 h-6" />
               </button>
-                </>
+            </div>
           )}
         </div>
       )}

src/components/playground/icons.tsx

@@ -184,7 +184,7 @@ export const SwitchCameraIcon = ({ className }: { className?: string }) => (
     strokeLinejoin="round"
     className={className}
   >
-    <path d="M20 4v5h-.58c-.89-3.1-3.85-5.23-7.17-5.23-4.26 0-7.74 3.4-7.74 7.6 0 .65.09 1.29.26 1.91l1.69-1.69c-.14-.22-.2-.46-.2-.72 0-3.2 2.63-5.81 5.87-5.81 2.67 0 4.96 1.75 5.6 4.23h-1.73l2.5 2.5 2.5-2.5h-1z" />
-    <path d="M4 20v-5h.58c.89 3.1 3.85 5.23 7.17 5.23 4.26 0 7.74-3.4 7.74-7.6 0-.65-.09-1.29-.26-1.91l-1.69 1.69c.14.22.2.46.2.72 0 3.2-2.63 5.81-5.87 5.81-2.67 0-4.96-1.75-5.6-4.23h1.73l-2.5-2.5-2.5 2.5h1z" />
+    <path d="M4 12v-3a3 3 0 0 1 3-3h13m-3-3 3 3-3 3" />
+    <path d="M20 12v3a3 3 0 0 1-3 3H4m3 3-3-3 3-3" />
   </svg>
 );