diff --git a/crates/ruma-common/src/events/audio.rs b/crates/ruma-common/src/events/audio.rs
index 68e1fd56..eb46b626 100644
--- a/crates/ruma-common/src/events/audio.rs
+++ b/crates/ruma-common/src/events/audio.rs
@@ -13,9 +13,25 @@ mod waveform_serde;
 
 use waveform_serde::WaveformSerDeHelper;
 
-use super::{file::FileContent, message::MessageContent, room::message::Relation};
+use super::{
+    file::FileContent,
+    message::MessageContent,
+    room::message::{AudioInfo, AudioMessageEventContent, Relation},
+};
 
 /// The payload for an extensible audio message.
+///
+/// This is the new primary type introduced in [MSC3246] and should not be sent before the end of
+/// the transition period. See the documentation of the [`message`] module for more information.
+///
+/// `AudioEventContent` can be converted to a [`RoomMessageEventContent`] with a
+/// [`MessageType::Audio`]. You can convert it back with
+/// [`AudioEventContent::from_audio_room_message()`].
+///
+/// [MSC3246]: https://github.com/matrix-org/matrix-spec-proposals/pull/3246
+/// [`message`]: super::message
+/// [`RoomMessageEventContent`]: super::room::message::RoomMessageEventContent
+/// [`MessageType::Audio`]: super::room::message::MessageType::Audio
 #[derive(Clone, Debug, Serialize, Deserialize, EventContent)]
 #[cfg_attr(not(feature = "unstable-exhaustive-types"), non_exhaustive)]
 #[ruma_event(type = "m.audio", kind = MessageLike)]
@@ -52,6 +68,23 @@ impl AudioEventContent {
     pub fn with_message(message: MessageContent, file: FileContent) -> Self {
         Self { message, file, audio: Default::default(), relates_to: None }
     }
+
+    /// Create a new `AudioEventContent` from the given `AudioMessageEventContent` and optional
+    /// relation.
+    pub fn from_audio_room_message(
+        content: AudioMessageEventContent,
+        relates_to: Option<Relation>,
+    ) -> Self {
+        let AudioMessageEventContent { body, source, info, message, file, audio } = content;
+
+        let message = message.unwrap_or_else(|| MessageContent::plain(body));
+        let file = file.unwrap_or_else(|| {
+            FileContent::from_room_message_content(source, info.as_deref(), None)
+        });
+        let audio = audio.or_else(|| info.as_deref().map(Into::into)).unwrap_or_default();
+
+        Self { message, file, audio, relates_to }
+    }
 }
 
 /// Audio content.
@@ -76,6 +109,18 @@ impl AudioContent {
     pub fn new() -> Self {
         Self::default()
     }
+
+    /// Whether this `AudioContent` is empty.
+    pub fn is_empty(&self) -> bool {
+        self.duration.is_none() && self.waveform.is_none()
+    }
+}
+
+impl From<&AudioInfo> for AudioContent {
+    fn from(info: &AudioInfo) -> Self {
+        let AudioInfo { duration, .. } = info;
+        Self { duration: duration.to_owned(), ..Default::default() }
+    }
 }
 
 /// The waveform representation of audio content.
diff --git a/crates/ruma-common/src/events/file.rs b/crates/ruma-common/src/events/file.rs
index ac3b5a25..5c0654ac 100644
--- a/crates/ruma-common/src/events/file.rs
+++ b/crates/ruma-common/src/events/file.rs
@@ -11,7 +11,7 @@ use serde::{Deserialize, Serialize};
 use super::{
     message::MessageContent,
     room::{
-        message::{FileInfo, FileMessageEventContent, Relation, VideoInfo},
+        message::{AudioInfo, FileInfo, FileMessageEventContent, Relation, VideoInfo},
         EncryptedFile, ImageInfo, JsonWebKey, MediaSource,
     },
 };
@@ -211,6 +211,13 @@ impl FileContentInfo {
     }
 }
 
+impl From<&AudioInfo> for FileContentInfo {
+    fn from(info: &AudioInfo) -> Self {
+        let AudioInfo { mimetype, size, .. } = info;
+        Self { mimetype: mimetype.to_owned(), size: size.to_owned(), ..Default::default() }
+    }
+}
+
 impl From<&FileInfo> for FileContentInfo {
     fn from(info: &FileInfo) -> Self {
         let FileInfo { mimetype, size, .. } = info;
diff --git a/crates/ruma-common/src/events/room/message.rs b/crates/ruma-common/src/events/room/message.rs
index 3bda3074..92434d2a 100644
--- a/crates/ruma-common/src/events/room/message.rs
+++ b/crates/ruma-common/src/events/room/message.rs
@@ -10,6 +10,8 @@ use serde::{de::DeserializeOwned, Deserialize, Serialize};
 use serde_json::Value as JsonValue;
 
 use super::{EncryptedFile, ImageInfo, MediaSource, ThumbnailInfo};
+#[cfg(feature = "unstable-msc3246")]
+use crate::events::audio::{AudioContent, AudioEventContent};
 #[cfg(feature = "unstable-msc3551")]
 use crate::events::file::{FileContent, FileContentInfo, FileEventContent};
 #[cfg(feature = "unstable-msc3552")]
@@ -194,6 +196,20 @@ impl RoomMessageEventContent {
     }
 }
 
+#[cfg(feature = "unstable-msc3246")]
+impl From<AudioEventContent> for RoomMessageEventContent {
+    fn from(content: AudioEventContent) -> Self {
+        let AudioEventContent { message, file, audio, relates_to } = content;
+
+        Self {
+            msgtype: MessageType::Audio(AudioMessageEventContent::from_extensible_content(
+                message, file, audio,
+            )),
+            relates_to,
+        }
+    }
+}
+
 #[cfg(feature = "unstable-msc1767")]
 impl From<EmoteEventContent> for RoomMessageEventContent {
     fn from(content: EmoteEventContent) -> Self {
@@ -518,6 +534,10 @@ impl Thread {
 #[derive(Clone, Debug, Deserialize, Serialize)]
 #[cfg_attr(not(feature = "unstable-exhaustive-types"), non_exhaustive)]
 #[serde(tag = "msgtype", rename = "m.audio")]
+#[cfg_attr(
+    feature = "unstable-msc3246",
+    serde(from = "content_serde::AudioMessageEventContentDeHelper")
+)]
 pub struct AudioMessageEventContent {
     /// The textual representation of this message.
     pub body: String,
@@ -529,19 +549,81 @@ pub struct AudioMessageEventContent {
     /// Metadata for the audio clip referred to in `url`.
     #[serde(skip_serializing_if = "Option::is_none")]
     pub info: Option<Box<AudioInfo>>,
+
+    /// Extensible-event text representation of the message.
+    ///
+    /// If present, this should be preferred over the `body` field.
+    #[cfg(feature = "unstable-msc3246")]
+    #[serde(flatten, skip_serializing_if = "Option::is_none")]
+    pub message: Option<MessageContent>,
+
+    /// Extensible-event file content of the message.
+    ///
+    /// If present, this should be preferred over the `source` and `info` fields.
+    #[cfg(feature = "unstable-msc3246")]
+    #[serde(rename = "org.matrix.msc1767.file", skip_serializing_if = "Option::is_none")]
+    pub file: Option<FileContent>,
+
+    /// Extensible-event audio info of the message.
+    ///
+    /// If present, this should be preferred over the `info` field.
+    #[cfg(feature = "unstable-msc3246")]
+    #[serde(rename = "org.matrix.msc1767.audio", skip_serializing_if = "Option::is_none")]
+    pub audio: Option<AudioContent>,
 }
 
 impl AudioMessageEventContent {
     /// Creates a new non-encrypted `RoomAudioMessageEventContent` with the given body, url and
     /// optional extra info.
     pub fn plain(body: String, url: Box<MxcUri>, info: Option<Box<AudioInfo>>) -> Self {
-        Self { body, source: MediaSource::Plain(url), info }
+        Self {
+            #[cfg(feature = "unstable-msc3246")]
+            message: Some(MessageContent::plain(body.clone())),
+            #[cfg(feature = "unstable-msc3246")]
+            file: Some(FileContent::plain(
+                url.clone(),
+                info.as_deref().map(|info| Box::new(info.into())),
+            )),
+            #[cfg(feature = "unstable-msc3246")]
+            audio: Some(info.as_deref().map_or_else(AudioContent::default, Into::into)),
+            body,
+            source: MediaSource::Plain(url),
+            info,
+        }
     }
 
     /// Creates a new encrypted `RoomAudioMessageEventContent` with the given body and encrypted
     /// file.
     pub fn encrypted(body: String, file: EncryptedFile) -> Self {
-        Self { body, source: MediaSource::Encrypted(Box::new(file)), info: None }
+        Self {
+            #[cfg(feature = "unstable-msc3246")]
+            message: Some(MessageContent::plain(body.clone())),
+            #[cfg(feature = "unstable-msc3246")]
+            file: Some(FileContent::encrypted(file.url.clone(), (&file).into(), None)),
+            #[cfg(feature = "unstable-msc3246")]
+            audio: Some(AudioContent::default()),
+            body,
+            source: MediaSource::Encrypted(Box::new(file)),
+            info: None,
+        }
+    }
+
+    /// Create a new `AudioMessageEventContent` with the given message, file info and audio info.
+    #[cfg(feature = "unstable-msc3246")]
+    pub fn from_extensible_content(
+        message: MessageContent,
+        file: FileContent,
+        audio: AudioContent,
+    ) -> Self {
+        let body = if let Some(body) = message.find_plain() {
+            body.to_owned()
+        } else {
+            message[0].body.clone()
+        };
+        let source = (&file).into();
+        let info = AudioInfo::from_extensible_content(file.info.as_deref(), &audio).map(Box::new);
+
+        Self { message: Some(message), file: Some(file), audio: Some(audio), body, source, info }
     }
 }
 
@@ -571,6 +653,24 @@ impl AudioInfo {
     pub fn new() -> Self {
         Self::default()
     }
+
+    /// Create an `AudioInfo` from the given file info and audio info.
+    #[cfg(feature = "unstable-msc3246")]
+    pub fn from_extensible_content(
+        file_info: Option<&FileContentInfo>,
+        audio: &AudioContent,
+    ) -> Option<Self> {
+        if file_info.is_none() && audio.is_empty() {
+            None
+        } else {
+            let (mimetype, size) = file_info
+                .map(|info| (info.mimetype.to_owned(), info.size.to_owned()))
+                .unwrap_or_default();
+            let AudioContent { duration, .. } = audio;
+
+            Some(Self { duration: duration.to_owned(), mimetype, size })
+        }
+    }
 }
 
 /// The payload for an emote message.
diff --git a/crates/ruma-common/src/events/room/message/content_serde.rs b/crates/ruma-common/src/events/room/message/content_serde.rs
index 8d8a6301..db491dfe 100644
--- a/crates/ruma-common/src/events/room/message/content_serde.rs
+++ b/crates/ruma-common/src/events/room/message/content_serde.rs
@@ -3,6 +3,8 @@
 use serde::{de, Deserialize};
 use serde_json::value::RawValue as RawJsonValue;
 
+#[cfg(feature = "unstable-msc3246")]
+use super::{AudioContent, AudioInfo, AudioMessageEventContent};
 #[cfg(feature = "unstable-msc3551")]
 use super::{FileContent, FileInfo, FileMessageEventContent, MediaSource, MessageContent};
 #[cfg(feature = "unstable-msc3552")]
@@ -57,6 +59,65 @@ impl<'de> Deserialize<'de> for MessageType {
     }
 }
 
+/// Helper struct for deserializing `AudioMessageEventContent` with stable and unstable field names.
+///
+/// It's not possible to use the `alias` attribute of serde because of
+/// https://github.com/serde-rs/serde/issues/1504.
+#[derive(Clone, Debug, Deserialize)]
+#[cfg(feature = "unstable-msc3246")]
+pub struct AudioMessageEventContentDeHelper {
+    /// The textual representation of this message.
+    pub body: String,
+
+    /// The source of the audio clip.
+    #[serde(flatten)]
+    pub source: MediaSource,
+
+    /// Metadata for the audio clip referred to in `source`.
+    pub info: Option<Box<AudioInfo>>,
+
+    /// Extensible-event text representation of the message.
+    #[serde(flatten)]
+    pub message: Option<MessageContent>,
+
+    /// Extensible-event file content of the message, with stable name.
+    #[serde(rename = "m.file")]
+    pub file_stable: Option<FileContent>,
+
+    /// Extensible-event file content of the message, with unstable name.
+    #[serde(rename = "org.matrix.msc1767.file")]
+    pub file_unstable: Option<FileContent>,
+
+    /// Extensible-event audio info of the message, with stable name.
+    #[serde(rename = "m.audio")]
+    pub audio_stable: Option<AudioContent>,
+
+    /// Extensible-event audio info of the message, with unstable name.
+    #[serde(rename = "org.matrix.msc1767.audio")]
+    pub audio_unstable: Option<AudioContent>,
+}
+
+#[cfg(feature = "unstable-msc3246")]
+impl From<AudioMessageEventContentDeHelper> for AudioMessageEventContent {
+    fn from(helper: AudioMessageEventContentDeHelper) -> Self {
+        let AudioMessageEventContentDeHelper {
+            body,
+            source,
+            info,
+            message,
+            file_stable,
+            file_unstable,
+            audio_stable,
+            audio_unstable,
+        } = helper;
+
+        let file = file_stable.or(file_unstable);
+        let audio = audio_stable.or(audio_unstable);
+
+        Self { body, source, info, message, file, audio }
+    }
+}
+
 /// Helper struct for deserializing `FileMessageEventContent` with stable and unstable field names.
 ///
 /// It's not possible to use the `alias` attribute of serde because of
diff --git a/crates/ruma-common/tests/events/audio.rs b/crates/ruma-common/tests/events/audio.rs
index 7c0778a8..c61dba1e 100644
--- a/crates/ruma-common/tests/events/audio.rs
+++ b/crates/ruma-common/tests/events/audio.rs
@@ -12,8 +12,10 @@ use ruma_common::{
         file::{EncryptedContentInit, FileContent, FileContentInfo},
         message::MessageContent,
         room::{
-            message::{InReplyTo, Relation},
-            JsonWebKeyInit,
+            message::{
+                AudioMessageEventContent, InReplyTo, MessageType, Relation, RoomMessageEventContent,
+            },
+            JsonWebKeyInit, MediaSource,
         },
         AnyMessageLikeEvent, MessageLikeEvent, MessageLikeUnsigned,
     },
@@ -208,7 +210,7 @@ fn event_serialization() {
 #[test]
 fn plain_content_deserialization() {
     let json_data = json!({
-        "org.matrix.msc1767.text": "Upload: my_new_song.webm",
+        "m.text": "Upload: my_new_song.webm",
         "m.file": {
             "url": "mxc://notareal.hs/abcdef",
         },
@@ -289,7 +291,7 @@ fn plain_content_deserialization() {
 #[test]
 fn encrypted_content_deserialization() {
     let json_data = json!({
-        "org.matrix.msc1767.text": "Upload: my_file.txt",
+        "m.text": "Upload: my_file.txt",
         "m.file": {
             "url": "mxc://notareal.hs/abcdef",
             "key": {
@@ -328,7 +330,7 @@ fn encrypted_content_deserialization() {
 fn message_event_deserialization() {
     let json_data = json!({
         "content": {
-            "org.matrix.msc1767.text": "Upload: airplane_sound.opus",
+            "m.text": "Upload: airplane_sound.opus",
             "m.file": {
                 "url": "mxc://notareal.hs/abcdef",
                 "name": "airplane_sound.opus",
@@ -379,3 +381,87 @@ fn message_event_deserialization() {
             && unsigned.is_empty()
     );
 }
+
+#[test]
+fn room_message_serialization() {
+    let message_event_content =
+        RoomMessageEventContent::new(MessageType::Audio(AudioMessageEventContent::plain(
+            "Upload: my_song.mp3".to_owned(),
+            mxc_uri!("mxc://notareal.hs/file").to_owned(),
+            None,
+        )));
+
+    assert_eq!(
+        to_json_value(&message_event_content).unwrap(),
+        json!({
+            "body": "Upload: my_song.mp3",
+            "url": "mxc://notareal.hs/file",
+            "msgtype": "m.audio",
+            "org.matrix.msc1767.text": "Upload: my_song.mp3",
+            "org.matrix.msc1767.file": {
+                "url": "mxc://notareal.hs/file",
+            },
+            "org.matrix.msc1767.audio": {},
+        })
+    );
+}
+
+#[test]
+fn room_message_stable_deserialization() {
+    let json_data = json!({
+        "body": "Upload: my_song.mp3",
+        "url": "mxc://notareal.hs/file",
+        "msgtype": "m.audio",
+        "m.text": "Upload: my_song.mp3",
+        "m.file": {
+            "url": "mxc://notareal.hs/file",
+        },
+        "m.audio": {},
+    });
+
+    let event_content = from_json_value::<RoomMessageEventContent>(json_data).unwrap();
+    assert_matches!(event_content.msgtype, MessageType::Audio(_));
+    if let MessageType::Audio(content) = event_content.msgtype {
+        assert_eq!(content.body, "Upload: my_song.mp3");
+        assert_matches!(content.source, MediaSource::Plain(_));
+        if let MediaSource::Plain(url) = content.source {
+            assert_eq!(url, "mxc://notareal.hs/file");
+        }
+        let message = content.message.unwrap();
+        assert_eq!(message.len(), 1);
+        assert_eq!(message[0].body, "Upload: my_song.mp3");
+        let file = content.file.unwrap();
+        assert_eq!(file.url, "mxc://notareal.hs/file");
+        assert!(!file.is_encrypted());
+    }
+}
+
+#[test]
+fn room_message_unstable_deserialization() {
+    let json_data = json!({
+        "body": "Upload: my_song.mp3",
+        "url": "mxc://notareal.hs/file",
+        "msgtype": "m.audio",
+        "org.matrix.msc1767.text": "Upload: my_song.mp3",
+        "org.matrix.msc1767.file": {
+            "url": "mxc://notareal.hs/file",
+        },
+        "org.matrix.msc1767.audio": {},
+    });
+
+    let event_content = from_json_value::<RoomMessageEventContent>(json_data).unwrap();
+    assert_matches!(event_content.msgtype, MessageType::Audio(_));
+    if let MessageType::Audio(content) = event_content.msgtype {
+        assert_eq!(content.body, "Upload: my_song.mp3");
+        assert_matches!(content.source, MediaSource::Plain(_));
+        if let MediaSource::Plain(url) = content.source {
+            assert_eq!(url, "mxc://notareal.hs/file");
+        }
+        let message = content.message.unwrap();
+        assert_eq!(message.len(), 1);
+        assert_eq!(message[0].body, "Upload: my_song.mp3");
+        let file = content.file.unwrap();
+        assert_eq!(file.url, "mxc://notareal.hs/file");
+        assert!(!file.is_encrypted());
+    }
+}
diff --git a/crates/ruma-common/tests/events/room_message.rs b/crates/ruma-common/tests/events/room_message.rs
index e12163d6..7203366f 100644
--- a/crates/ruma-common/tests/events/room_message.rs
+++ b/crates/ruma-common/tests/events/room_message.rs
@@ -52,6 +52,7 @@ fn serialization() {
         unsigned: MessageLikeUnsigned::default(),
     };
 
+    #[cfg(not(feature = "unstable-msc3246"))]
     assert_eq!(
         to_json_value(ev).unwrap(),
         json!({
@@ -67,6 +68,28 @@ fn serialization() {
             }
         })
     );
+
+    #[cfg(feature = "unstable-msc3246")]
+    assert_eq!(
+        to_json_value(ev).unwrap(),
+        json!({
+            "type": "m.room.message",
+            "event_id": "$143273582443PhrSn:example.org",
+            "origin_server_ts": 10_000,
+            "room_id": "!testroomid:example.org",
+            "sender": "@user:example.org",
+            "content": {
+                "body": "test",
+                "msgtype": "m.audio",
+                "url": "mxc://example.org/ffed755USFFxlgbQYZGtryd",
+                "org.matrix.msc1767.text": "test",
+                "org.matrix.msc1767.file": {
+                    "url": "mxc://example.org/ffed755USFFxlgbQYZGtryd",
+                },
+                "org.matrix.msc1767.audio": {},
+            }
+        })
+    );
 }
 
 #[test]
@@ -78,6 +101,7 @@ fn content_serialization() {
             None,
         )));
 
+    #[cfg(not(feature = "unstable-msc3246"))]
     assert_eq!(
         to_json_value(&message_event_content).unwrap(),
         json!({
@@ -86,6 +110,21 @@ fn content_serialization() {
             "url": "mxc://example.org/ffed755USFFxlgbQYZGtryd"
         })
     );
+
+    #[cfg(feature = "unstable-msc3246")]
+    assert_eq!(
+        to_json_value(&message_event_content).unwrap(),
+        json!({
+            "body": "test",
+            "msgtype": "m.audio",
+            "url": "mxc://example.org/ffed755USFFxlgbQYZGtryd",
+            "org.matrix.msc1767.text": "test",
+            "org.matrix.msc1767.file": {
+                "url": "mxc://example.org/ffed755USFFxlgbQYZGtryd",
+            },
+            "org.matrix.msc1767.audio": {},
+        })
+    );
 }
 
 #[test]