From 7fa52013653465661441ac016886ee843855a08c Mon Sep 17 00:00:00 2001
From: Ahmed Ibrahim <aibrahim@openai.com>
Date: Fri, 13 Mar 2026 16:17:13 -0700
Subject: [PATCH] Use parser-specific realtime voice enum (#14636)

Model realtime session output voices with an enum and map by parser so
v1 uses fathom and v2 uses alloy.

Co-authored-by: Codex <noreply@openai.com>
---
 .../src/endpoint/realtime_websocket/methods.rs | 18 ++++++++++--------
 .../endpoint/realtime_websocket/protocol.rs    | 10 +++++++++-
 2 files changed, 19 insertions(+), 9 deletions(-)

diff --git a/codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs b/codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs
index 12293693e..e62649e42 100644
--- a/codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs
+++ b/codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs
@@ -14,6 +14,7 @@ use crate::endpoint::realtime_websocket::protocol::SessionAudio;
 use crate::endpoint::realtime_websocket::protocol::SessionAudioFormat;
 use crate::endpoint::realtime_websocket::protocol::SessionAudioInput;
 use crate::endpoint::realtime_websocket::protocol::SessionAudioOutput;
+use crate::endpoint::realtime_websocket::protocol::SessionAudioVoice;
 use crate::endpoint::realtime_websocket::protocol::SessionFunctionTool;
 use crate::endpoint::realtime_websocket::protocol::SessionUpdateSession;
 use crate::endpoint::realtime_websocket::protocol::parse_realtime_event;
@@ -47,7 +48,6 @@ use tungstenite::protocol::WebSocketConfig;
 use url::Url;
 
 const REALTIME_AUDIO_SAMPLE_RATE: u32 = 24_000;
-const REALTIME_AUDIO_VOICE: &str = "fathom";
 const REALTIME_V1_SESSION_TYPE: &str = "quicksilver";
 const REALTIME_V2_SESSION_TYPE: &str = "realtime";
 const REALTIME_V2_CODEX_TOOL_NAME: &str = "codex";
@@ -353,13 +353,11 @@ impl RealtimeWebsocketWriter {
                     RealtimeEventParser::V1 => REALTIME_V1_SESSION_TYPE.to_string(),
                     RealtimeEventParser::RealtimeV2 => REALTIME_V2_SESSION_TYPE.to_string(),
                 };
-                (
-                    kind,
-                    Some(instructions),
-                    Some(SessionAudioOutput {
-                        voice: REALTIME_AUDIO_VOICE.to_string(),
-                    }),
-                )
+                let voice = match self.event_parser {
+                    RealtimeEventParser::V1 => SessionAudioVoice::Fathom,
+                    RealtimeEventParser::RealtimeV2 => SessionAudioVoice::Alloy,
+                };
+                (kind, Some(instructions), Some(SessionAudioOutput { voice }))
             }
             RealtimeSessionMode::Transcription => ("transcription".to_string(), None, None),
         };
@@ -1388,6 +1386,10 @@ mod tests {
                 first_json["session"]["type"],
                 Value::String("realtime".to_string())
             );
+            assert_eq!(
+                first_json["session"]["audio"]["output"]["voice"],
+                Value::String("alloy".to_string())
+            );
             assert_eq!(
                 first_json["session"]["tools"][0]["type"],
                 Value::String("function".to_string())
diff --git a/codex-rs/codex-api/src/endpoint/realtime_websocket/protocol.rs b/codex-rs/codex-api/src/endpoint/realtime_websocket/protocol.rs
index 028f51cf3..73c2c1052 100644
--- a/codex-rs/codex-api/src/endpoint/realtime_websocket/protocol.rs
+++ b/codex-rs/codex-api/src/endpoint/realtime_websocket/protocol.rs
@@ -77,7 +77,15 @@ pub(super) struct SessionAudioFormat {
 
 #[derive(Debug, Clone, Serialize)]
 pub(super) struct SessionAudioOutput {
-    pub(super) voice: String,
+    pub(super) voice: SessionAudioVoice,
+}
+
+#[derive(Debug, Clone, Copy, Serialize)]
+pub(super) enum SessionAudioVoice {
+    #[serde(rename = "fathom")]
+    Fathom,
+    #[serde(rename = "alloy")]
+    Alloy,
 }
 
 #[derive(Debug, Clone, Serialize)]