chore: refactor attachment media layout and insert menu organization

web/src/components/MemoEditor/Toolbar/InsertMenu.tsx

@@ -2,6 +2,7 @@ import { LatLng } from "leaflet";
 import { uniqBy } from "lodash-es";
 import {
   FileIcon,
+  ImageIcon,
   LinkIcon,
   LoaderIcon,
   type LucideIcon,
@@ -131,14 +132,22 @@ const InsertMenu = (props: InsertMenuProps) => {
     setMoreSubmenuOpen(false);
   }, [onToggleFocusMode]);
 
+  const handleMediaUploadClick = useCallback(() => {
+    handleUploadClick("image/*,video/*");
+  }, [handleUploadClick]);
+
+  const handleFileUploadClick = useCallback(() => {
+    handleUploadClick();
+  }, [handleUploadClick]);
+
   const menuItems = useMemo(
     () =>
       [
         {
-          key: "upload",
-          label: t("editor.insert-menu.upload-file"),
-          icon: FileIcon,
-          onClick: handleUploadClick,
+          key: "upload-media",
+          label: t("attachment-library.tabs.media"),
+          icon: ImageIcon,
+          onClick: handleMediaUploadClick,
         },
         {
           key: "record-audio",
@@ -146,6 +155,12 @@ const InsertMenu = (props: InsertMenuProps) => {
           icon: MicIcon,
           onClick: () => props.onAudioRecorderClick?.(),
         },
+        {
+          key: "upload-file",
+          label: t("common.file"),
+          icon: FileIcon,
+          onClick: handleFileUploadClick,
+        },
         {
           key: "link",
           label: t("editor.insert-menu.link-memo"),
@@ -159,7 +174,7 @@ const InsertMenu = (props: InsertMenuProps) => {
           onClick: handleLocationClick,
         },
       ] satisfies Array<{ key: string; label: string; icon: LucideIcon; onClick: () => void }>,
-    [handleLocationClick, handleOpenLinkDialog, handleUploadClick, props, t],
+    [handleFileUploadClick, handleLocationClick, handleMediaUploadClick, handleOpenLinkDialog, props, t],
   );
 
   return (
@@ -171,14 +186,14 @@ const InsertMenu = (props: InsertMenuProps) => {
           </Button>
         </DropdownMenuTrigger>
         <DropdownMenuContent align="start">
-          {menuItems.slice(0, 2).map((item) => (
+          {menuItems.slice(0, 3).map((item) => (
             <DropdownMenuItem key={item.key} onClick={item.onClick}>
               <item.icon className="w-4 h-4" />
               {item.label}
             </DropdownMenuItem>
           ))}
           <DropdownMenuSeparator />
-          {menuItems.slice(2).map((item) => (
+          {menuItems.slice(3).map((item) => (
             <DropdownMenuItem key={item.key} onClick={item.onClick}>
               <item.icon className="w-4 h-4" />
               {item.label}

web/src/components/MemoEditor/hooks/useFileUpload.ts

@@ -26,8 +26,13 @@ export const useFileUpload = (onFilesSelected: (localFiles: LocalFile[]) => void
     if (fileInputRef.current) fileInputRef.current.value = "";
   };
 
-  const handleUploadClick = () => {
-    fileInputRef.current?.click();
+  const handleUploadClick = (accept = "*") => {
+    if (!fileInputRef.current) {
+      return;
+    }
+
+    fileInputRef.current.accept = accept;
+    fileInputRef.current.click();
   };
 
   return {

web/src/components/MemoMetadata/Attachment/AttachmentListView.tsx

 import { DownloadIcon, FileIcon, PaperclipIcon, PlayIcon } from "lucide-react";
+import type { PropsWithChildren } from "react";
 import { useMemo } from "react";
 import MetadataSection from "@/components/MemoMetadata/MetadataSection";
 import MotionPhotoPreview from "@/components/MotionPhotoPreview";
 import { cn } from "@/lib/utils";
 import type { Attachment } from "@/types/proto/api/v1/attachment_service_pb";
 import { getAttachmentUrl } from "@/utils/attachment";
-import type { PreviewMediaItem } from "@/utils/media-item";
+import type { AttachmentVisualItem, PreviewMediaItem } from "@/utils/media-item";
 import { buildAttachmentVisualItems } from "@/utils/media-item";
 import AudioAttachmentItem from "./AudioAttachmentItem";
 import { getAttachmentMetadata, isAudioAttachment, separateAttachments } from "./attachmentHelpers";
@@ -15,6 +16,17 @@ interface AttachmentListViewProps {
   onImagePreview?: (items: PreviewMediaItem[], index: number) => void;
 }
 
+type VisualItem = AttachmentVisualItem;
+
+const VISUAL_TILE_CLASS =
+  "group relative overflow-hidden rounded-xl border border-border/70 bg-muted/30 text-left transition-colors hover:border-accent/40";
+const COVER_MEDIA_CLASS = "h-full w-full rounded-none object-cover transition-transform duration-300 group-hover:scale-[1.02]";
+const NATURAL_MEDIA_CLASS =
+  "block h-auto max-h-[20rem] w-auto max-w-full rounded-none transition-transform duration-300 group-hover:scale-[1.02]";
+const SINGLE_VIDEO_CARD_WIDTH_CLASS = "w-full max-w-[30rem]";
+const TWO_ITEM_GRID_HEIGHT_CLASS = "h-[11rem] sm:h-[13rem] md:h-[15rem]";
+const MOSAIC_GRID_HEIGHT_CLASS = "h-[13rem] sm:h-[16rem] md:h-[18rem]";
+
 const AttachmentMeta = ({ attachment }: { attachment: Attachment }) => {
   const { fileTypeLabel, fileSizeLabel } = getAttachmentMetadata(attachment);
 
@@ -50,89 +62,175 @@ const DocumentItem = ({ attachment }: { attachment: Attachment }) => {
   );
 };
 
-const MotionItem = ({
+const getMotionPreviewProps = (item: VisualItem) => ({
+  motionUrl: item.previewItem.kind === "motion" ? item.previewItem.motionUrl : item.sourceUrl,
+  presentationTimestampUs: item.previewItem.kind === "motion" ? item.previewItem.presentationTimestampUs : undefined,
+});
+
+const VisualTile = ({
+  className,
+  onPreview,
+  overlayLabel,
+  children,
+}: PropsWithChildren<{ className?: string; onPreview?: () => void; overlayLabel?: string }>) => {
+  return (
+    <button type="button" className={cn(VISUAL_TILE_CLASS, className)} onClick={onPreview}>
+      {children}
+      <div className="pointer-events-none absolute inset-0 bg-gradient-to-t from-foreground/15 via-transparent to-transparent opacity-0 transition-opacity group-hover:opacity-100" />
+      {overlayLabel && (
+        <div className="pointer-events-none absolute inset-0 flex items-center justify-center bg-black/45 text-2xl font-semibold text-white backdrop-blur-[2px]">
+          {overlayLabel}
+        </div>
+      )}
+    </button>
+  );
+};
+
+const VideoPlayBadge = ({ className, children }: PropsWithChildren<{ className?: string }>) => (
+  <span
+    className={cn(
+      "pointer-events-none absolute inline-flex items-center justify-center rounded-full bg-background/85 text-foreground shadow-sm backdrop-blur-sm",
+      className,
+    )}
+  >
+    {children}
+  </span>
+);
+
+const CollageVisualItem = ({
   item,
-  featured = false,
   onPreview,
+  className,
+  overlayLabel,
 }: {
-  item: ReturnType<typeof buildAttachmentVisualItems>[number];
-  featured?: boolean;
+  item: VisualItem;
   onPreview?: () => void;
+  className?: string;
+  overlayLabel?: string;
 }) => {
+  const motionPreviewProps = item.kind === "motion" ? getMotionPreviewProps(item) : undefined;
+
   return (
-    <button
-      type="button"
-      className={cn("group block w-full text-left", featured ? "max-w-[18rem] sm:max-w-[20rem]" : "")}
-      onClick={onPreview}
-    >
-      <div
-        className={cn(
-          "relative overflow-hidden rounded-xl border border-border/70 bg-muted/30 transition-colors hover:border-accent/40",
-          featured ? "aspect-[4/3]" : "aspect-square",
-        )}
-      >
-        {item.kind === "video" ? (
-          <video
-            src={item.sourceUrl}
-            className="h-full w-full rounded-none object-cover transition-transform duration-300 group-hover:scale-[1.02]"
-            preload="metadata"
-          />
-        ) : item.kind === "motion" ? (
-          <MotionPhotoPreview
-            posterUrl={item.posterUrl}
-            motionUrl={item.previewItem.kind === "motion" ? item.previewItem.motionUrl : item.sourceUrl}
-            alt={item.filename}
-            presentationTimestampUs={item.previewItem.kind === "motion" ? item.previewItem.presentationTimestampUs : undefined}
-            containerClassName="h-full w-full"
-            badgeClassName="left-2 top-2 px-2 py-0.5 text-[10px]"
-            mediaClassName="h-full w-full rounded-none object-cover transition-transform duration-300 group-hover:scale-[1.02]"
-          />
-        ) : (
-          <img
-            src={item.posterUrl}
-            alt={item.filename}
-            className="h-full w-full rounded-none object-cover transition-transform duration-300 group-hover:scale-[1.02]"
-            loading="lazy"
-            decoding="async"
-          />
-        )}
-        <div className="pointer-events-none absolute inset-0 bg-gradient-to-t from-foreground/15 via-transparent to-transparent opacity-0 transition-opacity group-hover:opacity-100" />
-        {item.kind === "video" && (
-          <span className="pointer-events-none absolute bottom-2 right-2 inline-flex h-7 w-7 items-center justify-center rounded-full bg-background/80 text-foreground/70 backdrop-blur-sm">
-            <PlayIcon className="h-3.5 w-3.5 fill-current" />
-          </span>
-        )}
+    <VisualTile className={cn("block h-full w-full", className)} onPreview={onPreview} overlayLabel={overlayLabel}>
+      {item.kind === "video" ? (
+        <>
+          <video src={item.sourceUrl} className={COVER_MEDIA_CLASS} preload="metadata" />
+          {!overlayLabel && (
+            <VideoPlayBadge className="bottom-2 right-2 h-7 w-7 bg-background/80 text-foreground/70">
+              <PlayIcon className="h-3.5 w-3.5 fill-current" />
+            </VideoPlayBadge>
+          )}
+        </>
+      ) : item.kind === "motion" && motionPreviewProps ? (
+        <MotionPhotoPreview
+          posterUrl={item.posterUrl}
+          motionUrl={motionPreviewProps.motionUrl}
+          alt={item.filename}
+          presentationTimestampUs={motionPreviewProps.presentationTimestampUs}
+          containerClassName="h-full w-full"
+          badgeClassName="left-2 top-2 px-2 py-0.5 text-[10px]"
+          mediaClassName={COVER_MEDIA_CLASS}
+        />
+      ) : (
+        <img src={item.posterUrl} alt={item.filename} className={COVER_MEDIA_CLASS} loading="lazy" decoding="async" />
+      )}
+    </VisualTile>
+  );
+};
+
+const SingleVisualItem = ({ item, onPreview }: { item: VisualItem; onPreview?: () => void }) => {
+  const motionPreviewProps = item.kind === "motion" ? getMotionPreviewProps(item) : undefined;
+
+  if (item.kind === "image") {
+    return (
+      <VisualTile className="inline-block max-w-full" onPreview={onPreview}>
+        <img src={item.posterUrl} alt={item.filename} className={NATURAL_MEDIA_CLASS} loading="lazy" decoding="async" />
+      </VisualTile>
+    );
+  }
+
+  if (item.kind === "motion" && motionPreviewProps) {
+    return (
+      <VisualTile className="inline-block max-w-full" onPreview={onPreview}>
+        <MotionPhotoPreview
+          posterUrl={item.posterUrl}
+          motionUrl={motionPreviewProps.motionUrl}
+          alt={item.filename}
+          presentationTimestampUs={motionPreviewProps.presentationTimestampUs}
+          containerClassName="max-w-full"
+          posterClassName={cn(NATURAL_MEDIA_CLASS, "object-contain")}
+          videoClassName="absolute inset-0 h-full w-full rounded-none object-contain transition-transform duration-300 group-hover:scale-[1.02]"
+          badgeClassName="left-2 top-2 px-2 py-0.5 text-[10px]"
+        />
+      </VisualTile>
+    );
+  }
+
+  return (
+    <VisualTile className={cn("block", SINGLE_VIDEO_CARD_WIDTH_CLASS)} onPreview={onPreview}>
+      <div className="relative aspect-video bg-black/5">
+        <video src={item.sourceUrl} poster={item.posterUrl} className={COVER_MEDIA_CLASS} preload="metadata" />
+        <div className="pointer-events-none absolute inset-0 bg-gradient-to-t from-black/35 via-black/5 to-transparent" />
+        <VideoPlayBadge className="bottom-3 right-3 h-9 w-9">
+          <PlayIcon className="h-4 w-4 fill-current" />
+        </VideoPlayBadge>
       </div>
-    </button>
+    </VisualTile>
   );
 };
 
-const VisualGallery = ({
-  items,
-  onPreview,
-}: {
-  items: ReturnType<typeof buildAttachmentVisualItems>;
-  onPreview?: (itemId: string) => void;
-}) => {
+const VisualGallery = ({ items, onPreview }: { items: VisualItem[]; onPreview?: (itemId: string) => void }) => {
+  if (items.length === 0) {
+    return null;
+  }
+
   if (items.length === 1) {
     return (
-      <div className="flex">
-        <MotionItem item={items[0]} featured onPreview={() => onPreview?.(items[0].id)} />
+      <div className="w-full">
+        <SingleVisualItem item={items[0]} onPreview={() => onPreview?.(items[0].id)} />
+      </div>
+    );
+  }
+
+  if (items.length === 2) {
+    return (
+      <div className={cn("grid grid-cols-2 gap-2", TWO_ITEM_GRID_HEIGHT_CLASS)}>
+        {items.map((item) => (
+          <CollageVisualItem key={item.id} item={item} onPreview={() => onPreview?.(item.id)} />
+        ))}
       </div>
     );
   }
 
+  if (items.length === 3) {
+    return (
+      <div className={cn("grid grid-cols-2 grid-rows-2 gap-2", MOSAIC_GRID_HEIGHT_CLASS)}>
+        <CollageVisualItem item={items[0]} className="row-span-2" onPreview={() => onPreview?.(items[0].id)} />
+        <CollageVisualItem item={items[1]} onPreview={() => onPreview?.(items[1].id)} />
+        <CollageVisualItem item={items[2]} onPreview={() => onPreview?.(items[2].id)} />
+      </div>
+    );
+  }
+
+  const visibleItems = items.slice(0, 4);
+  const remainingCount = items.length - visibleItems.length;
+
   return (
-    <div className="grid max-w-[22rem] grid-cols-2 gap-1.5 sm:max-w-[24rem]">
-      {items.map((item) => (
-        <MotionItem key={item.id} item={item} onPreview={() => onPreview?.(item.id)} />
+    <div className={cn("grid grid-cols-2 grid-rows-2 gap-2", MOSAIC_GRID_HEIGHT_CLASS)}>
+      {visibleItems.map((item, index) => (
+        <CollageVisualItem
+          key={item.id}
+          item={item}
+          overlayLabel={index === visibleItems.length - 1 && remainingCount > 0 ? `+${remainingCount}` : undefined}
+          onPreview={() => onPreview?.(item.id)}
+        />
       ))}
     </div>
   );
 };
 
-const AudioList = ({ attachments }: { attachments: Attachment[] }) => (
-  <div className="flex flex-col gap-2">
+const AudioList = ({ attachments, compact = false }: { attachments: Attachment[]; compact?: boolean }) => (
+  <div className={cn("gap-2", compact ? "grid grid-cols-1 sm:grid-cols-2" : "flex flex-col")}>
     {attachments.map((attachment) => (
       <AudioAttachmentItem
         key={attachment.name}
@@ -140,6 +238,7 @@ const AudioList = ({ attachments }: { attachments: Attachment[] }) => (
         sourceUrl={getAttachmentUrl(attachment)}
         mimeType={attachment.type}
         size={Number(attachment.size)}
+        compact={compact}
       />
     ))}
   </div>
@@ -164,7 +263,7 @@ const AttachmentListView = ({ attachments, onImagePreview }: AttachmentListViewP
   const hasVisual = visualItems.length > 0;
   const hasAudio = audio.length > 0;
   const hasDocs = docs.length > 0;
-  const sectionCount = [hasVisual, hasAudio, hasDocs].filter(Boolean).length;
+  const hasMedia = hasVisual || hasAudio;
 
   if (attachments.length === 0) {
     return null;
@@ -182,10 +281,13 @@ const AttachmentListView = ({ attachments, onImagePreview }: AttachmentListViewP
       count={visualItems.length + audio.length + docs.length}
       contentClassName="flex flex-col gap-2 p-2"
     >
-      {hasVisual && <VisualGallery items={visualItems} onPreview={handlePreview} />}
-      {hasVisual && sectionCount > 1 && <Divider />}
-      {hasAudio && <AudioList attachments={audio.filter(isAudioAttachment)} />}
-      {hasAudio && hasDocs && <Divider />}
+      {hasMedia && (
+        <div className="flex flex-col gap-2">
+          {hasVisual && <VisualGallery items={visualItems} onPreview={handlePreview} />}
+          {hasAudio && <AudioList attachments={audio.filter(isAudioAttachment)} compact />}
+        </div>
+      )}
+      {hasMedia && hasDocs && <Divider />}
       {hasDocs && <DocsList attachments={docs} />}
     </MetadataSection>
   );

web/src/components/MemoMetadata/Attachment/AudioAttachmentItem.tsx

 import { PauseIcon, PlayIcon } from "lucide-react";
 import { useEffect, useRef, useState } from "react";
+import { cn } from "@/lib/utils";
 import { formatFileSize, getFileTypeLabel } from "@/utils/format";
 import { formatAudioTime } from "./attachmentHelpers";
 
@@ -56,9 +57,11 @@ interface AudioAttachmentItemProps {
   mimeType: string;
   size?: number;
   title?: string;
+  compact?: boolean;
+  className?: string;
 }
 
-const AudioAttachmentItem = ({ filename, sourceUrl, mimeType, size, title }: AudioAttachmentItemProps) => {
+const AudioAttachmentItem = ({ filename, sourceUrl, mimeType, size, title, compact = false, className }: AudioAttachmentItemProps) => {
   const audioRef = useRef<HTMLAudioElement>(null);
   const [isPlaying, setIsPlaying] = useState(false);
   const [currentTime, setCurrentTime] = useState(0);
@@ -119,9 +122,9 @@ const AudioAttachmentItem = ({ filename, sourceUrl, mimeType, size, title }: Aud
   };
 
   return (
-    <div className="rounded-xl border border-border/40 bg-background/75 px-2 py-1.5">
-      <div className="flex items-center justify-between gap-1.5">
-        <div className="min-w-0 flex flex-1 items-baseline gap-1">
+    <div className={cn("rounded-xl border border-border/40 bg-background/75", compact ? "px-3 py-2.5" : "px-2 py-1.5", className)}>
+      <div className="flex items-start justify-between gap-2">
+        <div className={cn("min-w-0 flex flex-1", compact ? "flex-col gap-0.5" : "items-baseline gap-1")}>
           <div className="truncate text-sm font-medium leading-5 text-foreground" title={filename}>
             {displayTitle}
           </div>
@@ -141,7 +144,7 @@ const AudioAttachmentItem = ({ filename, sourceUrl, mimeType, size, title }: Aud
         </button>
       </div>
 
-      <div className="mt-1 flex items-center gap-1">
+      <div className={cn("mt-1", compact ? "space-y-1.5" : "flex items-center gap-1")}>
         <AudioProgressBar
           filename={filename}
           currentTime={currentTime}
@@ -151,16 +154,18 @@ const AudioAttachmentItem = ({ filename, sourceUrl, mimeType, size, title }: Aud
           className="min-w-0 flex-1"
         />
 
-        <div className="shrink-0 text-[10px] tabular-nums text-muted-foreground">{timeLabel}</div>
-
-        <button
-          type="button"
-          onClick={handlePlaybackRateChange}
-          className="inline-flex h-5 shrink-0 items-center justify-center rounded-md border border-transparent px-1 text-[10px] font-medium text-muted-foreground transition-colors hover:border-border/40 hover:text-foreground"
-          aria-label={`Playback speed ${playbackRate}x for ${displayTitle}`}
-        >
-          {playbackRate}x
-        </button>
+        <div className={cn("flex items-center", compact ? "justify-between gap-2" : "shrink-0 gap-1")}>
+          <div className="shrink-0 text-[10px] tabular-nums text-muted-foreground">{timeLabel}</div>
+
+          <button
+            type="button"
+            onClick={handlePlaybackRateChange}
+            className="inline-flex h-5 shrink-0 items-center justify-center rounded-md border border-transparent px-1 text-[10px] font-medium text-muted-foreground transition-colors hover:border-border/40 hover:text-foreground"
+            aria-label={`Playback speed ${playbackRate}x for ${displayTitle}`}
+          >
+            {playbackRate}x
+          </button>
+        </div>
       </div>
 
       <audio

web/src/components/MotionPhotoPlayer.tsx

@@ -8,6 +8,8 @@ interface MotionPhotoPlayerProps {
   presentationTimestampUs?: bigint;
   containerClassName?: string;
   mediaClassName?: string;
+  posterClassName?: string;
+  videoClassName?: string;
   active?: boolean;
   loop?: boolean;
 }
@@ -19,6 +21,8 @@ const MotionPhotoPlayer = ({
   presentationTimestampUs,
   containerClassName,
   mediaClassName,
+  posterClassName,
+  videoClassName,
   active,
   loop = false,
 }: MotionPhotoPlayerProps) => {
@@ -87,7 +91,7 @@ const MotionPhotoPlayer = ({
       <img
         src={posterUrl}
         alt={alt}
-        className={cn("block max-h-full max-w-full select-none object-cover", mediaClassName)}
+        className={cn("block max-h-full max-w-full select-none object-cover", mediaClassName, posterClassName)}
         draggable={false}
         loading="lazy"
         decoding="async"
@@ -100,6 +104,7 @@ const MotionPhotoPlayer = ({
           "pointer-events-none absolute inset-0 h-full w-full object-cover transition-opacity duration-200",
           isPlaying ? "opacity-100" : "opacity-0",
           mediaClassName,
+          videoClassName,
         )}
         muted
         playsInline

web/src/components/MotionPhotoPreview.tsx

@@ -9,6 +9,8 @@ interface MotionPhotoPreviewProps {
   presentationTimestampUs?: bigint;
   containerClassName?: string;
   mediaClassName?: string;
+  posterClassName?: string;
+  videoClassName?: string;
   badgeClassName?: string;
   loop?: boolean;
 }
@@ -20,6 +22,8 @@ const MotionPhotoPreview = ({
   presentationTimestampUs,
   containerClassName,
   mediaClassName,
+  posterClassName,
+  videoClassName,
   badgeClassName,
   loop = false,
 }: MotionPhotoPreviewProps) => {
@@ -40,6 +44,8 @@ const MotionPhotoPreview = ({
         loop={loop}
         containerClassName={cn("max-w-full max-h-full", containerClassName)}
         mediaClassName={mediaClassName}
+        posterClassName={posterClassName}
+        videoClassName={videoClassName}
       />
       <div
         role="button"