diff --git a/crates/ruma-events/CHANGELOG.md b/crates/ruma-events/CHANGELOG.md
index 683bbc6d..48879d4a 100644
--- a/crates/ruma-events/CHANGELOG.md
+++ b/crates/ruma-events/CHANGELOG.md
@@ -23,6 +23,7 @@ The new format (Session) is required to reliably display the call member count (
 `CallMemberEventContent` is now an enum to model the two different formats.
 - `CallMemberStateKey` (instead of `OwnedUserId`) is now used as the state key type for `CallMemberEventContent`.
 This guarantees correct formatting of the event key.
+- Add helpers for captions on audio, file, image and video messages.
 
 Breaking changes:
 
diff --git a/crates/ruma-events/src/room/message.rs b/crates/ruma-events/src/room/message.rs
index 39f7475a..2f1d016e 100644
--- a/crates/ruma-events/src/room/message.rs
+++ b/crates/ruma-events/src/room/message.rs
@@ -30,6 +30,7 @@ mod file;
 mod image;
 mod key_verification_request;
 mod location;
+mod media_caption;
 mod notice;
 mod relation;
 pub(crate) mod relation_serde;
diff --git a/crates/ruma-events/src/room/message/audio.rs b/crates/ruma-events/src/room/message/audio.rs
index edd59f7e..438e3f5d 100644
--- a/crates/ruma-events/src/room/message/audio.rs
+++ b/crates/ruma-events/src/room/message/audio.rs
@@ -5,7 +5,10 @@ use ruma_common::OwnedMxcUri;
 use serde::{Deserialize, Serialize};
 
 use super::FormattedBody;
-use crate::room::{EncryptedFile, MediaSource};
+use crate::room::{
+    message::media_caption::{caption, formatted_caption},
+    EncryptedFile, MediaSource,
+};
 
 /// The payload for an audio message.
 #[derive(Clone, Debug, Deserialize, Serialize)]
@@ -88,6 +91,21 @@ impl AudioMessageEventContent {
     pub fn info(self, info: impl Into<Option<Box<AudioInfo>>>) -> Self {
         Self { info: info.into(), ..self }
     }
+
+    /// Returns the caption for the audio as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// In short, this is the `body` field if the `filename` field exists and has a different value,
+    /// otherwise the media file does not have a caption.
+    pub fn caption(&self) -> Option<&str> {
+        caption(&self.body, self.filename.as_deref())
+    }
+
+    /// Returns the formatted caption for the audio as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// This is the same as `caption`, but returns the formatted body instead of the plain body.
+    pub fn formatted_caption(&self) -> Option<&FormattedBody> {
+        formatted_caption(&self.body, self.formatted.as_ref(), self.filename.as_deref())
+    }
 }
 
 /// Metadata about an audio clip.
diff --git a/crates/ruma-events/src/room/message/file.rs b/crates/ruma-events/src/room/message/file.rs
index 3f7abb51..0edbc634 100644
--- a/crates/ruma-events/src/room/message/file.rs
+++ b/crates/ruma-events/src/room/message/file.rs
@@ -3,7 +3,10 @@ use ruma_common::OwnedMxcUri;
 use serde::{Deserialize, Serialize};
 
 use super::FormattedBody;
-use crate::room::{EncryptedFile, MediaSource, ThumbnailInfo};
+use crate::room::{
+    message::media_caption::{caption, formatted_caption},
+    EncryptedFile, MediaSource, ThumbnailInfo,
+};
 
 /// The payload for a file message.
 #[derive(Clone, Debug, Deserialize, Serialize)]
@@ -60,6 +63,21 @@ impl FileMessageEventContent {
     pub fn info(self, info: impl Into<Option<Box<FileInfo>>>) -> Self {
         Self { info: info.into(), ..self }
     }
+
+    /// Returns the caption of the media file as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// In short, this is the `body` field if the `filename` field exists and has a different value,
+    /// otherwise the media file does not have a caption.
+    pub fn caption(&self) -> Option<&str> {
+        caption(&self.body, self.filename.as_deref())
+    }
+
+    /// Returns the formatted caption of the media file as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// This is the same as `caption`, but returns the formatted body instead of the plain body.
+    pub fn formatted_caption(&self) -> Option<&FormattedBody> {
+        formatted_caption(&self.body, self.formatted.as_ref(), self.filename.as_deref())
+    }
 }
 
 /// Metadata about a file.
diff --git a/crates/ruma-events/src/room/message/image.rs b/crates/ruma-events/src/room/message/image.rs
index 674be973..29e1c60c 100644
--- a/crates/ruma-events/src/room/message/image.rs
+++ b/crates/ruma-events/src/room/message/image.rs
@@ -2,7 +2,10 @@ use ruma_common::OwnedMxcUri;
 use serde::{Deserialize, Serialize};
 
 use super::FormattedBody;
-use crate::room::{EncryptedFile, ImageInfo, MediaSource};
+use crate::room::{
+    message::media_caption::{caption, formatted_caption},
+    EncryptedFile, ImageInfo, MediaSource,
+};
 
 /// The payload for an image message.
 #[derive(Clone, Debug, Deserialize, Serialize)]
@@ -59,4 +62,19 @@ impl ImageMessageEventContent {
     pub fn info(self, info: impl Into<Option<Box<ImageInfo>>>) -> Self {
         Self { info: info.into(), ..self }
     }
+
+    /// Returns the caption for the image as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// In short, this is the `body` field if the `filename` field exists and has a different value,
+    /// otherwise the media file does not have a caption.
+    pub fn caption(&self) -> Option<&str> {
+        caption(&self.body, self.filename.as_deref())
+    }
+
+    /// Returns the formatted caption for the image as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// This is the same as `caption`, but returns the formatted body instead of the plain body.
+    pub fn formatted_caption(&self) -> Option<&FormattedBody> {
+        formatted_caption(&self.body, self.formatted.as_ref(), self.filename.as_deref())
+    }
 }
diff --git a/crates/ruma-events/src/room/message/media_caption.rs b/crates/ruma-events/src/room/message/media_caption.rs
new file mode 100644
index 00000000..a24a688e
--- /dev/null
+++ b/crates/ruma-events/src/room/message/media_caption.rs
@@ -0,0 +1,22 @@
+//! Reusable methods for captioning media files.
+
+use crate::room::message::FormattedBody;
+
+/// Computes the caption of a media file as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+///
+/// In short, this is the `body` field if the `filename` field exists and has a different value,
+/// otherwise the media file does not have a caption.
+pub(crate) fn caption<'a>(body: &'a str, filename: Option<&str>) -> Option<&'a str> {
+    filename.is_some_and(|filename| body != filename).then_some(body)
+}
+
+/// Computes the formatted caption of a media file as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+///
+/// This is the same as `caption`, but returns the formatted body instead of the plain body.
+pub(crate) fn formatted_caption<'a>(
+    body: &str,
+    formatted: Option<&'a FormattedBody>,
+    filename: Option<&str>,
+) -> Option<&'a FormattedBody> {
+    filename.is_some_and(|filename| body != filename).then_some(formatted).flatten()
+}
diff --git a/crates/ruma-events/src/room/message/video.rs b/crates/ruma-events/src/room/message/video.rs
index 00d9dd00..df8ac0c1 100644
--- a/crates/ruma-events/src/room/message/video.rs
+++ b/crates/ruma-events/src/room/message/video.rs
@@ -5,7 +5,10 @@ use ruma_common::OwnedMxcUri;
 use serde::{Deserialize, Serialize};
 
 use super::FormattedBody;
-use crate::room::{EncryptedFile, MediaSource, ThumbnailInfo};
+use crate::room::{
+    message::media_caption::{caption, formatted_caption},
+    EncryptedFile, MediaSource, ThumbnailInfo,
+};
 
 /// The payload for a video message.
 #[derive(Clone, Debug, Deserialize, Serialize)]
@@ -62,6 +65,21 @@ impl VideoMessageEventContent {
     pub fn info(self, info: impl Into<Option<Box<VideoInfo>>>) -> Self {
         Self { info: info.into(), ..self }
     }
+
+    /// Returns the caption of the video as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// In short, this is the `body` field if the `filename` field exists and has a different value,
+    /// otherwise the media file does not have a caption.
+    pub fn caption(&self) -> Option<&str> {
+        caption(&self.body, self.filename.as_deref())
+    }
+
+    /// Returns the formatted caption of the video as defined by the [spec](https://spec.matrix.org/latest/client-server-api/#media-captions).
+    ///
+    /// This is the same as `caption`, but returns the formatted body instead of the plain body.
+    pub fn formatted_caption(&self) -> Option<&FormattedBody> {
+        formatted_caption(&self.body, self.formatted.as_ref(), self.filename.as_deref())
+    }
 }
 
 /// Metadata about a video.
diff --git a/crates/ruma-events/tests/it/room_message.rs b/crates/ruma-events/tests/it/room_message.rs
index acc2422b..717f5685 100644
--- a/crates/ruma-events/tests/it/room_message.rs
+++ b/crates/ruma-events/tests/it/room_message.rs
@@ -12,7 +12,7 @@ use ruma_events::{
     room::{
         message::{
             AddMentions, AudioMessageEventContent, EmoteMessageEventContent,
-            FileMessageEventContent, ForwardThread, ImageMessageEventContent,
+            FileMessageEventContent, FormattedBody, ForwardThread, ImageMessageEventContent,
             KeyVerificationRequestEventContent, MessageType, OriginalRoomMessageEvent,
             OriginalSyncRoomMessageEvent, Relation, ReplyWithinThread, RoomMessageEventContent,
             TextMessageEventContent, VideoMessageEventContent,
@@ -900,8 +900,9 @@ fn audio_msgtype_deserialization() {
     let event_content = from_json_value::<RoomMessageEventContent>(json_data).unwrap();
     assert_matches!(event_content.msgtype, MessageType::Audio(content));
     assert_eq!(content.body, "Upload: my_song.mp3");
-    assert_matches!(content.source, MediaSource::Plain(url));
+    assert_matches!(&content.source, MediaSource::Plain(url));
     assert_eq!(url, "mxc://notareal.hs/file");
+    assert!(content.caption().is_none());
 }
 
 #[test]
@@ -983,8 +984,9 @@ fn file_msgtype_plain_content_deserialization() {
     let event_content = from_json_value::<RoomMessageEventContent>(json_data).unwrap();
     assert_matches!(event_content.msgtype, MessageType::File(content));
     assert_eq!(content.body, "Upload: my_file.txt");
-    assert_matches!(content.source, MediaSource::Plain(url));
+    assert_matches!(&content.source, MediaSource::Plain(url));
     assert_eq!(url, "mxc://notareal.hs/file");
+    assert!(content.caption().is_none());
 }
 
 #[test]
@@ -1045,8 +1047,9 @@ fn image_msgtype_deserialization() {
     let event_content = from_json_value::<RoomMessageEventContent>(json_data).unwrap();
     assert_matches!(event_content.msgtype, MessageType::Image(content));
     assert_eq!(content.body, "Upload: my_image.jpg");
-    assert_matches!(content.source, MediaSource::Plain(url));
+    assert_matches!(&content.source, MediaSource::Plain(url));
     assert_eq!(url, "mxc://notareal.hs/file");
+    assert!(content.caption().is_none());
 }
 
 #[cfg(not(feature = "unstable-msc3488"))]
@@ -1202,8 +1205,9 @@ fn video_msgtype_deserialization() {
     let event_content = from_json_value::<RoomMessageEventContent>(json_data).unwrap();
     assert_matches!(event_content.msgtype, MessageType::Video(content));
     assert_eq!(content.body, "Upload: my_video.mp4");
-    assert_matches!(content.source, MediaSource::Plain(url));
+    assert_matches!(&content.source, MediaSource::Plain(url));
     assert_eq!(url, "mxc://notareal.hs/file");
+    assert!(content.caption().is_none());
 }
 
 #[test]
@@ -1330,3 +1334,106 @@ fn invalid_replacement() {
     assert_matches!(&data, Cow::Borrowed(_)); // data is stored in JSON form because it's invalid
     assert_eq!(JsonValue::Object(data.into_owned()), relation);
 }
+
+#[test]
+fn test_audio_caption() {
+    let mut content = AudioMessageEventContent::plain(
+        "my_sound.ogg".to_owned(),
+        mxc_uri!("mxc://notareal.hs/abcdef").to_owned(),
+    );
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.filename = Some("my_sound.ogg".to_owned());
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.body = "This was a great podcast episode".to_owned();
+    assert_eq!(content.caption(), Some("This was a great podcast episode"));
+    assert!(content.formatted_caption().is_none());
+
+    content.formatted =
+        Some(FormattedBody::html("This was a <em>great</em> podcast episode".to_owned()));
+    assert_eq!(content.caption(), Some("This was a great podcast episode"));
+    assert_eq!(
+        content.formatted_caption().map(|f| f.body.clone()),
+        Some("This was a <em>great</em> podcast episode".to_owned())
+    );
+}
+
+#[test]
+fn test_file_caption() {
+    let mut content = FileMessageEventContent::plain(
+        "my_file.txt".to_owned(),
+        mxc_uri!("mxc://notareal.hs/abcdef").to_owned(),
+    );
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.filename = Some("my_file.txt".to_owned());
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.body = "Please check these notes".to_owned();
+    assert_eq!(content.caption(), Some("Please check these notes"));
+    assert!(content.formatted_caption().is_none());
+
+    content.formatted =
+        Some(FormattedBody::html("<strong>Please check these notes</strong>".to_owned()));
+    assert_eq!(content.caption(), Some("Please check these notes"));
+    assert_eq!(
+        content.formatted_caption().map(|f| f.body.clone()),
+        Some("<strong>Please check these notes</strong>".to_owned())
+    );
+}
+
+#[test]
+fn test_image_caption() {
+    let mut content = ImageMessageEventContent::plain(
+        "my_image.jpg".to_owned(),
+        mxc_uri!("mxc://notareal.hs/abcdef").to_owned(),
+    );
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.filename = Some("my_image.jpg".to_owned());
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.body = "Check it out 😎".to_owned();
+    assert_eq!(content.caption(), Some("Check it out 😎"));
+    assert!(content.formatted_caption().is_none());
+
+    content.formatted = Some(FormattedBody::html("<h3>Check it out 😎</h3>".to_owned()));
+    assert_eq!(content.caption(), Some("Check it out 😎"));
+    assert_eq!(
+        content.formatted_caption().map(|f| f.body.clone()),
+        Some("<h3>Check it out 😎</h3>".to_owned())
+    );
+}
+
+#[test]
+fn test_video_caption() {
+    let mut content = VideoMessageEventContent::plain(
+        "my_video.mp4".to_owned(),
+        mxc_uri!("mxc://notareal.hs/abcdef").to_owned(),
+    );
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.filename = Some("my_video.mp4".to_owned());
+    assert!(content.caption().is_none());
+    assert!(content.formatted_caption().is_none());
+
+    content.body = "You missed a great evening".to_owned();
+    assert_eq!(content.caption(), Some("You missed a great evening"));
+    assert!(content.formatted_caption().is_none());
+
+    content.formatted =
+        Some(FormattedBody::html("You missed a <strong>great</strong> evening".to_owned()));
+    assert_eq!(content.caption(), Some("You missed a great evening"));
+    assert_eq!(
+        content.formatted_caption().map(|f| f.body.clone()),
+        Some("You missed a <strong>great</strong> evening".to_owned())
+    );
+}