diff --git a/desktop/CHANGELOG.md b/desktop/CHANGELOG.md
index 091ace6c38..34699edc79 100644
--- a/desktop/CHANGELOG.md
+++ b/desktop/CHANGELOG.md
@@ -2,6 +2,7 @@
 
 ## v1.7.8 (Unreleased)
 
+- Parse description from image metadata.
 - .
 
 ## v1.7.7
diff --git a/web/apps/photos/src/services/upload/upload-service.ts b/web/apps/photos/src/services/upload/upload-service.ts
index 66ebe28406..38ce924081 100644
--- a/web/apps/photos/src/services/upload/upload-service.ts
+++ b/web/apps/photos/src/services/upload/upload-service.ts
@@ -1073,7 +1073,8 @@ const extractImageOrVideoMetadata = async (
         if (h) publicMagicMetadata.h = ensureInteger(h);
     }
 
-    const caption = parsedMetadataJSON?.description;
+    const caption =
+        parsedMetadataJSON?.description ?? parsedMetadata.description;
     if (caption) {
         publicMagicMetadata.caption = caption;
     }
diff --git a/web/packages/media/file-metadata.ts b/web/packages/media/file-metadata.ts
index 75bc8bf6c0..b431a24f76 100644
--- a/web/packages/media/file-metadata.ts
+++ b/web/packages/media/file-metadata.ts
@@ -604,6 +604,10 @@ export interface ParsedMetadata {
     creationDate?: ParsedMetadataDate;
     /** The GPS coordinates where the photo was taken. */
     location?: Location;
+    /**
+     * A caption / description attached by the user to the photo.
+     */
+    description?: string;
 }
 
 /**
diff --git a/web/packages/new/photos/services/exif.ts b/web/packages/new/photos/services/exif.ts
index 782434323e..5c76398327 100644
--- a/web/packages/new/photos/services/exif.ts
+++ b/web/packages/new/photos/services/exif.ts
@@ -40,10 +40,12 @@ export const parseExif = (tags: RawExifTags) => {
     const location = parseLocation(tags);
     const creationDate = parseCreationDate(tags);
     const dimensions = parseDimensions(tags);
+    const description = parseDescription(tags);
 
     const metadata: ParsedMetadata = dimensions ?? {};
     if (creationDate) metadata.creationDate = creationDate;
     if (location) metadata.location = location;
+    if (description) metadata.description = description;
     return metadata;
 };
 
@@ -536,3 +538,25 @@ export const tagNumericValue = (
     const v = tag.value;
     return Array.isArray(v) ? (v[0] ?? 0) / (v[1] ?? 1) : v;
 };
+
+/**
+ * Parse the description for an image from the metadata embedded in the file.
+ *
+ * This function will read the description from the following fields, in order:
+ *
+ * 1. XMP-dc:description
+ * 2. IPTC | Caption/Abstract (120)
+ * 3. EXIF IFD0 | 0x010e | ImageDescription
+ *
+ * For an overview of why this ordering was chosen, see
+ * https://github.com/ente-io/ente/discussions/3857#discussioncomment-11764990
+ */
+const parseDescription = (tags: RawExifTags) =>
+    // While the TypeScript tags for these three fields are varying (The XMP one
+    // doesn't have a static type, the IPTC one is marked as a number array, and
+    // the Exif one is a string array), for all three of these, the ExifReader
+    // description property (not related to the image "description" in the sense
+    // of this function) holds the singular string we're interested in.
+    tags.xmp?.description?.description ??
+    tags.iptc?.["Caption/Abstract"]?.description ??
+    tags.exif?.ImageDescription?.description;