diff --git a/client/builders.go b/client/builders.go
index be40c32a..2b0ce1b3 100644
--- a/client/builders.go
+++ b/client/builders.go
@@ -386,7 +386,10 @@ func (c *QQClient) buildDeleteOnlinePushPacket(uin int64, seq uint16, delMsg []j
 func (c *QQClient) buildGroupSendingPacket(groupCode int64, r int32, forward bool, m *message.SendingMessage) (uint16, []byte) {
 	seq := c.nextSeq()
 	var ptt *message.GroupVoiceElement
-	if i := m.FirstOrNil(func(e message.IMessageElement) bool { return e.Type() == message.Voice }); i != nil {
+	if i := m.FirstOrNil(func(e message.IMessageElement) bool {
+		_, ok := e.(*message.GroupVoiceElement)
+		return ok
+	}); i != nil {
 		ptt = i.(*message.GroupVoiceElement)
 		m.Elements = []message.IMessageElement{}
 	}
diff --git a/client/global.go b/client/global.go
index 878d08e7..81acd80d 100644
--- a/client/global.go
+++ b/client/global.go
@@ -269,6 +269,16 @@ func (c *QQClient) parseGroupMessage(m *msg.Message) *message.GroupMessage {
 		Elements:  message.ParseMessageElems(m.Body.RichText.Elems),
 		//OriginalElements: m.Body.RichText.Elems,
 	}
+	if m.Body.RichText.Ptt != nil {
+		g.Elements = []message.IMessageElement{
+			&message.VoiceElement{
+				Name:         m.Body.RichText.Ptt.FileName,
+				Md5:          m.Body.RichText.Ptt.FileMd5,
+				Size:         m.Body.RichText.Ptt.FileSize,
+				DownloadPara: string(m.Body.RichText.Ptt.DownPara),
+			},
+		}
+	}
 	if m.Body.RichText.Attr != nil {
 		g.InternalId = m.Body.RichText.Attr.Random
 	}
diff --git a/message/elements.go b/message/elements.go
index cfcc7d8b..a9139b63 100644
--- a/message/elements.go
+++ b/message/elements.go
@@ -25,6 +25,16 @@ type GroupImageElement struct {
 	Url     string
 }
 
+type VoiceElement struct {
+	Name         string
+	Md5          []byte
+	Size         int32
+	DownloadPara string
+
+	// --- sending ---
+	Data []byte
+}
+
 type GroupVoiceElement struct {
 	Data []byte
 	Ptt  *msg.Ptt
@@ -186,6 +196,10 @@ func (e *GroupVoiceElement) Type() ElementType {
 	return Voice
 }
 
+func (e *VoiceElement) Type() ElementType {
+	return Voice
+}
+
 var faceMap = map[int]string{
 	14:  "微笑",
 	1:   "撇嘴",