Merge pull request #88 from sloganking/codex/add-cli-flag-for-push-to-talk-audio-ducking

Add push-to-talk audio ducking option

Author: sloganking

Cargo.lock

@@ -153,7 +153,7 @@ impl DefaultDeviceSink {
             let speed = inner.sink.speed();
             let paused = inner.sink.is_paused();
 
-            let (stream, mut new_sink, name) = Self::create_stream();
+            let (stream, new_sink, name) = Self::create_stream();
             new_sink.set_volume(volume);
             new_sink.set_speed(speed);
 

src/default_device_sink.rs

@@ -53,11 +53,10 @@ use uuid::Uuid;
 mod options;
 #[cfg(target_os = "windows")]
 mod windows_volume;
+use speakstream::ss::SpeakStream;
 use tracing::{debug, error, info, instrument, warn};
 use tracing_appender::rolling::{RollingFileAppender, Rotation};
 
-use speakstream::ss::SpeakStream;
-
 #[derive(Debug, Subcommand)]
 pub enum SubCommands {
     /// Displays keys as you press them so you can figure out what key to use for push to talk.
@@ -1072,6 +1071,7 @@ async fn main() -> Result<(), Box<dyn Error>> {
             // It then sends the path of the recorded audio file to the AI thread.
             let thread_llm_should_stop_mutex = llm_should_stop_mutex.clone();
             let thread_speak_stream_mutex = speak_stream_mutex.clone();
+            let duck_ptt = opt.duck_ptt;
             thread::spawn(move || {
                 let mut recorder = rec::Recorder::new();
                 let mut recording_start = std::time::SystemTime::now();
@@ -1100,6 +1100,9 @@ async fn main() -> Result<(), Box<dyn Error>> {
                                     let mut thread_speak_stream =
                                         thread_speak_stream_mutex.lock().unwrap();
                                     thread_speak_stream.stop_speech();
+                                    if duck_ptt {
+                                        thread_speak_stream.start_audio_ducking();
+                                    }
                                     drop(thread_speak_stream);
                                 }
 
@@ -1149,6 +1152,13 @@ async fn main() -> Result<(), Box<dyn Error>> {
                                     }
                                 }
 
+                                if duck_ptt {
+                                    let mut thread_speak_stream =
+                                        thread_speak_stream_mutex.lock().unwrap();
+                                    thread_speak_stream.stop_audio_ducking();
+                                    drop(thread_speak_stream);
+                                }
+
                                 // continue if we failed to get elapsed time
                                 let elapsed = match elapsed_option {
                                     Some(elapsed) => elapsed,

src/main.rs

@@ -1,6 +1,5 @@
 use crate::{SubCommands, VoiceEnum};
 use clap::Parser;
-use easy_rdev_key::PTTKey;
 
 #[derive(Parser, Debug)]
 #[command(version)]
@@ -33,6 +32,10 @@ pub struct Opt {
     #[arg(long)]
     pub tick: bool,
 
+    /// Duck other application audio while the push-to-talk key is held down.
+    #[arg(long)]
+    pub duck_ptt: bool,
+
     /// Start with the AI voice muted.
     #[arg(long)]
     pub mute: bool,