gui: Add terminator packet and 200ms voice hold for VAD
Implements proper voice activity detection with: - 200ms hold period after audio drops below threshold to prevent choppy cutoffs - Terminator packet (end_bit=true) when speech ends to signal stream completion - TransmitState enum to track transmission state across frames This ensures other Mumble clients receive proper end-of-speech signaling for clean audio termination and correct "talking" indicator behavior. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
+29
-16
@@ -1,4 +1,4 @@
|
||||
use crate::effects::{AudioProcessor, AudioProcessorSender};
|
||||
use crate::effects::{AudioProcessor, AudioProcessorSender, TransmitState};
|
||||
use color_eyre::eyre::{eyre, Error};
|
||||
use cpal::traits::{DeviceTrait, HostTrait, StreamTrait as _};
|
||||
use futures::io::{AsyncRead, AsyncWrite};
|
||||
@@ -23,6 +23,31 @@ pub struct AudioSystem {
|
||||
const SAMPLE_RATE: u32 = 48_000;
|
||||
const PACKET_SAMPLES: u32 = 960;
|
||||
|
||||
fn encode_and_send(
|
||||
state: TransmitState,
|
||||
output_buffer: &mut Vec<f32>,
|
||||
encoder: &mut opus::Encoder,
|
||||
each: &mut impl FnMut(Vec<u8>, bool),
|
||||
) {
|
||||
let (is_terminator, should_encode) = match state {
|
||||
TransmitState::Silent => return,
|
||||
TransmitState::Transmitting => (false, output_buffer.len() >= PACKET_SAMPLES as usize),
|
||||
TransmitState::Terminator => {
|
||||
output_buffer.resize(PACKET_SAMPLES as usize, 0.0);
|
||||
(true, true)
|
||||
}
|
||||
};
|
||||
|
||||
if should_encode {
|
||||
let remainder = output_buffer.split_off(PACKET_SAMPLES as usize);
|
||||
let frame = replace(output_buffer, remainder);
|
||||
match encoder.encode_vec_float(&frame, frame.len() * 2) {
|
||||
Ok(encoded) => each(encoded, is_terminator),
|
||||
Err(e) => error!("error encoding {} samples: {e:?}", frame.len()),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
type Buffer = Arc<Mutex<dasp_ring_buffer::Bounded<Vec<i16>>>>;
|
||||
|
||||
impl AudioSystem {
|
||||
@@ -79,7 +104,7 @@ impl AudioSystem {
|
||||
|
||||
pub fn start_recording(
|
||||
&mut self,
|
||||
mut each: impl FnMut(Vec<u8>) + Send + 'static,
|
||||
mut each: impl FnMut(Vec<u8>, bool) + Send + 'static,
|
||||
) -> Result<(), Error> {
|
||||
let config = self.choose_config(self.input.supported_input_configs()?)?;
|
||||
info!(
|
||||
@@ -97,20 +122,8 @@ impl AudioSystem {
|
||||
if let Some(new_processor) = processors.take() {
|
||||
current_processor = new_processor;
|
||||
}
|
||||
current_processor.process(frame, config.channels as usize, &mut output_buffer);
|
||||
if output_buffer.len() < PACKET_SAMPLES as usize {
|
||||
return;
|
||||
}
|
||||
let remainder = output_buffer.split_off(PACKET_SAMPLES as usize);
|
||||
let frame = replace(&mut output_buffer, remainder);
|
||||
match encoder.encode_vec_float(&frame, frame.len() * 2) {
|
||||
Ok(buf) => {
|
||||
each(buf);
|
||||
}
|
||||
Err(e) => {
|
||||
error!("error encoding {} samples: {e:?}", frame.len());
|
||||
}
|
||||
}
|
||||
let state = current_processor.process(frame, config.channels as usize, &mut output_buffer);
|
||||
encode_and_send(state, &mut output_buffer, &mut encoder, &mut each);
|
||||
};
|
||||
|
||||
match self
|
||||
|
||||
Reference in New Issue
Block a user