fix audio ringbuf and improve vad

2 files changed, 36 insertions, 19 deletions

diff --git a/client/src/audio.rs b/client/src/audio.rs
index f93466d..d90898a 100644
--- a/client/src/audio.rs
+++ b/client/src/audio.rs
@@ -56,7 +56,7 @@ impl Audio {
         inconf.buffer_size = cpal::BufferSize::Fixed(480);
 
         let mut outconf = outdev.default_input_config()?.config();
-        outconf.channels = 1;
+        outconf.channels = 2;
         outconf.sample_rate = cpal::SampleRate(48_000);
         outconf.buffer_size = cpal::BufferSize::Fixed(480);
 
@@ -101,6 +101,7 @@ impl Audio {
     }
 }
 
+#[derive(Debug)]
 pub struct APlayPacket {
     pos: Vec3,
     channel: u128,
@@ -156,10 +157,11 @@ impl AEncoder {
             let (tx, end_tx) = self.trigger.update(energy);
             if tx {
                 let size = self.encoder.encode_float(&denoise, &mut out)?;
+                debug!("encoded size={size}");
                 let _ = self.sender.try_send(out[..size].to_vec());
             }
             if end_tx {
-                // TODO send end frame
+                let _ = self.sender.try_send(vec![]);
             }
         }
         Ok(())
@@ -168,14 +170,14 @@ impl AEncoder {
 
 impl VadTrigger {
     pub fn update(&mut self, energy: f32) -> (bool, bool) {
-        debug!("E={energy:.02}");
         let now = Instant::now();
-        if energy > 1. {
+        let thres = if self.transmitting { 0.1 } else { 1. };
+        if energy > thres {
             self.last_sig = now;
         }
         let last_sig_elapsed = (now - self.last_sig).as_secs_f32();
         let prev_transmitting = self.transmitting;
-        self.transmitting = last_sig_elapsed < 0.5;
+        self.transmitting = last_sig_elapsed < 0.2;
 
         match (prev_transmitting, self.transmitting) {
             (false, true) => info!("start transmit"),
@@ -188,6 +190,8 @@ impl VadTrigger {
 }
 
 const BUFFER_SIZE: usize = 48_000;
+const JITTER_COMP: usize = 48_00 * 2;
+
 pub struct ADecoder {
     decoder: Decoder,
     receiver: Receiver<APlayPacket>,
@@ -210,8 +214,12 @@ impl ADecoder {
         ))
     }
     pub fn data(&mut self, samples: &mut [f32]) -> Result<()> {
-        while self.buffer.len() < samples.len() {
-            if let Ok(p) = self.receiver.try_recv() {
+        for p in self.receiver.try_iter() {
+            if p.data.is_empty() {
+                if self.channels.remove(&p.channel).is_some() {
+                    info!("channel {} ended", p.channel % 1_000_000)
+                }
+            } else {
                 let mut output = [0f32; AE_FRAME_SIZE];
                 let size = self.decoder.decode_float(
                     Some(p.data.as_slice()),
@@ -219,17 +227,19 @@ impl ADecoder {
                     false,
                 )?;
 
-                let channel_cursor = self.channels.entry(p.channel).or_insert(self.playback);
+                let channel_cursor = self.channels.entry(p.channel).or_insert_with(|| {
+                    info!("channel {} started", p.channel % 1_000_000);
+                    (self.playback + JITTER_COMP) % BUFFER_SIZE
+                });
                 let free_space = *channel_cursor - self.playback;
                 for i in 0..size.min(free_space) {
                     // TODO positional audio
                     let _ = p.pos;
-                    self.buffer[*channel_cursor] = [output[i], output[i]];
+                    self.buffer[*channel_cursor][0] += output[i];
+                    self.buffer[*channel_cursor][1] += output[i];
                     *channel_cursor += 1;
                     *channel_cursor %= BUFFER_SIZE
                 }
-            } else {
-                break;
             }
         }
         for x in samples.array_chunks_mut::<2>() {
diff --git a/client/src/state.rs b/client/src/state.rs
index e59691a..560c8bd 100644
--- a/client/src/state.rs
+++ b/client/src/state.rs
@@ -41,6 +41,7 @@ pub struct State<'a> {
     pub prefab_index_res_loaded: Option<Resource<PrefabIndex>>,
     pub prefab_index_res: Option<Resource<PrefabIndex>>,
     pub avatar_ob: Option<Object>,
+    pub last_position_tx: Instant,
 }
 
 pub struct InputState {
@@ -73,6 +74,7 @@ impl<'a> State<'a> {
             prefab_index_res_loaded: None,
             avatar_ob: None,
             prefab_index: PrefabIndex::default().into(),
+            last_position_tx: Instant::now(),
         })
     }
     pub fn draw(&mut self) {
@@ -146,9 +148,11 @@ impl<'a> State<'a> {
             match &p {
                 Packet::PrefabIndex(res) => self.prefab_index_res = Some(res.to_owned()),
                 Packet::Sound(ob, data) => {
-                    if let Some(obdata) = self.tree.objects.get(ob) {
-                        self.audio
-                            .incoming_packet(ob.0, obdata.pos.into(), data.0.clone())
+                    if Some(ob) != self.avatar_ob.as_ref() {
+                        if let Some(obdata) = self.tree.objects.get(ob) {
+                            self.audio
+                                .incoming_packet(ob.0, obdata.pos.into(), data.0.clone())
+                        }
                     }
                 }
                 _ => (),
@@ -169,11 +173,14 @@ impl<'a> State<'a> {
         }
 
         if let Some(ob) = self.avatar_ob {
-            self.network.packet_send.send(Packet::Position(
-                ob,
-                self.camera.position().into(),
-                self.camera.rotation().into(),
-            ))?;
+            if self.last_position_tx.elapsed().as_secs_f32() > 0.1 {
+                self.network.packet_send.send(Packet::Position(
+                    ob,
+                    self.camera.position().into(),
+                    self.camera.rotation().into(),
+                ))?;
+                self.last_position_tx = Instant::now();
+            }
             while let Some(p) = self.audio.pop_output() {
                 self.network.packet_send.send(Packet::Sound(ob, Data(p)))?;
             }