rvsttd: single faded audio stream feeds both Moonshine and debug WAV
This commit is contained in:
+25
-21
@@ -113,6 +113,14 @@ impl DebugRecorder {
|
||||
self.audio.extend_from_slice(samples);
|
||||
}
|
||||
|
||||
fn audio_len(&self) -> usize {
|
||||
self.audio.len()
|
||||
}
|
||||
|
||||
fn truncate_audio(&mut self, len: usize) {
|
||||
self.audio.truncate(len);
|
||||
}
|
||||
|
||||
fn save(self) {
|
||||
// Write log
|
||||
let log_path = self.dir.join("session.log");
|
||||
@@ -325,48 +333,44 @@ fn transcriber_loop(
|
||||
// Fixed 100ms window — cpal delivers every ~50ms (800 samples @ 16kHz),
|
||||
// so this captures 1-2 more callbacks worth of trailing audio.
|
||||
let drain_deadline = std::time::Instant::now() + Duration::from_millis(100);
|
||||
let mut last_chunk: Option<Vec<f32>> = None;
|
||||
let mut drained: Vec<f32> = Vec::new();
|
||||
while std::time::Instant::now() < drain_deadline {
|
||||
match rx.recv_timeout(drain_deadline - std::time::Instant::now()) {
|
||||
Ok(chunk) => {
|
||||
// Record raw audio for debug WAV
|
||||
if let Some(ref mut r) = recorder {
|
||||
r.add_audio(&chunk);
|
||||
}
|
||||
|
||||
// Feed previous chunk to Moonshine, hold the latest for fade
|
||||
if let Some(prev) = last_chunk.take() {
|
||||
unsafe {
|
||||
moonshine_transcribe_add_audio_to_stream(
|
||||
handle, stream_handle,
|
||||
prev.as_ptr(), prev.len() as u64,
|
||||
SAMPLE_RATE, 0,
|
||||
);
|
||||
}
|
||||
}
|
||||
last_chunk = Some(chunk);
|
||||
drained.extend_from_slice(&chunk);
|
||||
}
|
||||
Err(mpsc::RecvTimeoutError::Timeout) => break,
|
||||
Err(mpsc::RecvTimeoutError::Disconnected) => break,
|
||||
}
|
||||
}
|
||||
|
||||
// Apply 5ms fade-out to the last chunk before feeding to Moonshine.
|
||||
// The recorder already captured the unmodified audio for the debug WAV.
|
||||
if let Some(mut chunk) = last_chunk.take() {
|
||||
// Apply 5ms fade-out to the trailing edge of the drained audio.
|
||||
// Both the debug WAV and Moonshine receive the same faded audio.
|
||||
let fade_samples = (SAMPLE_RATE as usize * 5) / 1000;
|
||||
if chunk.len() > fade_samples {
|
||||
let start = chunk.len() - fade_samples;
|
||||
if drained.len() > fade_samples {
|
||||
let start = drained.len() - fade_samples;
|
||||
for i in 0..fade_samples {
|
||||
let t = 1.0 - (i as f32 / fade_samples as f32);
|
||||
chunk[start + i] *= t;
|
||||
drained[start + i] *= t;
|
||||
}
|
||||
}
|
||||
|
||||
// Update recorder with the faded audio (replaces the raw versions added above)
|
||||
if let Some(ref mut r) = recorder {
|
||||
// Remove the raw drained samples and re-add the faded ones
|
||||
let raw_drained_len = r.audio_len();
|
||||
r.truncate_audio(raw_drained_len.saturating_sub(drained.len()));
|
||||
r.add_audio(&drained);
|
||||
}
|
||||
|
||||
if !drained.is_empty() {
|
||||
unsafe {
|
||||
moonshine_transcribe_add_audio_to_stream(
|
||||
handle, stream_handle,
|
||||
chunk.as_ptr(), chunk.len() as u64,
|
||||
drained.as_ptr(), drained.len() as u64,
|
||||
SAMPLE_RATE, 0,
|
||||
);
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user