diff --git a/crates/captions/src/vtt.rs b/crates/captions/src/vtt.rs
index 8c1d57bd2..a1afb2445 100644
--- a/crates/captions/src/vtt.rs
+++ b/crates/captions/src/vtt.rs
@@ -38,7 +38,7 @@ pub fn split_voice(text: &str) -> (Option<String>, String) {
             Some(_) => return (None, text.to_string()),
             None => body.to_string(),
         };
-        if body.contains("<v ") || body.contains("<v.") {
+        if body.contains("<v ") || body.contains("<v.") || body.contains("<v\t") {
             return (None, text.to_string());
         }
         return ((!name.is_empty()).then(|| name.to_string()), body);
@@ -177,6 +177,19 @@ mod tests {
         assert_eq!(split_voice("<vx>hi"), (None, "<vx>hi".into()));
     }
 
+    #[test]
+    fn multiple_voices_with_tab_separator() {
+        let text = "<v Ann>Hello <v\tBob>there";
+        assert_eq!(split_voice(text), (None, text.to_string()));
+        let source = format!("WEBVTT\n\n00:00:01.000 --> 00:00:02.000\n{text}\n");
+        let doc = parse(&source).unwrap();
+        assert_eq!(doc.cues[0].speaker, None);
+        assert_eq!(doc.cues[0].text, text);
+        assert_eq!(parse(&write(&doc)).unwrap().cues, doc.cues);
+        let closed = "<v Ann>Hi <v\tBob>x</v>";
+        assert_eq!(split_voice(closed), (None, closed.to_string()));
+    }
+
     #[test]
     fn without_signature() {
         let d = parse("00:00:01.000 --> 00:00:02.000\nx\n").unwrap();