gramps/LlamaCasty
0
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix(audio): drive the mix from the clock — the audio track was 2.1% longer than the show

Browse Source 
Measured on the creator's recording (ty-20261112-1118-423-0000-2.mp4), which drifts
out of sync over ten seconds:

  wall clock (pump start -> audio stop)   9.875 s
  video: 594 frames, uniform 1/60         9.888 s   correct
  audio: 485,376 samples @ 48 kHz        10.112 s   +2.1%, 224 ms long

So it was not VFR and not wall-clock PTS: the video is CFR and matches the clock (the
only PTS anomaly in the file is one 38 ms gap from two lost frames). The audio track is
what was too long, and a static offset cannot fix a rate error — which is why the opening
looked aligned (the 175 ms mic delay cancels the camera's constant pipeline latency) and
the end did not.

Cause: LiveLoopAsync wrote a fixed 480 frames per iteration and trusted that each
iteration was 10 ms. Under a raised timer resolution they land near 9.8 ms — 102 ticks
per second where 100 are expected, i.e. 102% of an audio second per wall second. FFmpeg
stamps audio by sample count, so every second of show pushed the sound ~2% later against
a correct picture. Video never had the bug because the pump paces off a stopwatch.

Fix: Services/Audio/AudioClock asks the clock how many frames are owed
(expected - produced, clamped to one 200 ms chunk) and the loop produces exactly that,
padding on underrun. A stalled loop pays its debt over the following ticks instead of
dumping one huge write. FillAndMix takes a frame count, and the master limiter/meter take
spans, so a partial chunk is processed without touching stale buffer tail.

674 tests (one log-rotation flake under full-suite load while the app itself is running
and writing startup.log; passes alone), 0 warnings. Ear check pending.
This commit is contained in:
gramps
2026-11-12 11:28:28 -08:00
parent 7f89534309
commit 47f409f0e4
5 changed files with 156 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Services/Audio/AudioClock.cs
+34
View File
@@ -0,0 +1,34 @@
using System;
namespace ytLive.Services.Audio;
/// <summary>
/// How many frames the mix must produce <i>now</i>, derived from the monotonic clock rather than from a
/// tick count.
///
/// <para><b>Why this exists (2026-11-12):</b> the live loop used to write a fixed 480 frames per
/// iteration and trust that every iteration was 10 ms. It is not: `Task.Delay` plus processing lands
/// near 9.8 ms under a raised timer resolution, and 102 ticks per second where 100 are expected is
/// 102% of an audio second per wall second. FFmpeg stamps audio by sample count, so the track comes out
/// longer than the show and the sound slides later against a correct picture — measured on a 9.9 s
/// recording: video 9.888 s (right), audio 10.112 s (+2.1%, 224 ms).</para>
///
/// <para>The video side never had this bug because the pump paces off a stopwatch. This is the audio
/// equivalent: ask the clock what should exist, produce the difference, pad on underrun. Pure, so the
/// drift case is a test rather than a listening test.</para>
/// </summary>
public static class AudioClock
{
/// <summary>
/// Frames still owed at <paramref name="elapsedSeconds"/>: expected-minus-produced, clamped to
/// 0..<paramref name="maxFrames"/>. Zero means "not due yet"; the cap keeps a stalled loop from
/// dumping a huge burst in one write (the deficit stays owed and is paid off over the next ticks).
/// </summary>
public static int FramesDue(double elapsedSeconds, long framesProduced, int sampleRate, int maxFrames)
{
if (elapsedSeconds <= 0 || sampleRate <= 0 || maxFrames <= 0) return 0;
var expected = (long)Math.Round(elapsedSeconds * sampleRate);
var debt = expected - framesProduced;
return (int)Math.Clamp(debt, 0L, maxFrames);
}
}
Services/Audio/AudioMixer.cs
+62 -36
View File
@@ -1,3 +1,4 @@
using System.Diagnostics;
using ytLive.Helpers;
using ytLive.Services;
@@ -121,16 +122,19 @@ public sealed class AudioMixer : IDisposable
var seconds = _mixInterval.TotalSeconds;
var framesPerTick = Math.Max(1, (int)Math.Round(OutputSampleRate * seconds));
// A tick normally owes ~10 ms; after a stall it may owe more, and the clock-based loop pays it
// off in one write. Buffers therefore size to the largest chunk the planner can return.
var maxChunkFrames = Math.Max(framesPerTick, OutputSampleRate / 5);
_micBuffer = new AudioRingBuffer(OutputSampleRate * 2);
_loopbackBuffer = new AudioRingBuffer(OutputSampleRate * 2 * 2);
_alertBuffer = new AudioRingBuffer(OutputSampleRate * 2 * 8);
_voiceChain = new VoiceFilterChain(OutputSampleRate);
_ducker = new AutoDucker();
_micChunk = new float[framesPerTick];
_loopbackChunk = new float[framesPerTick * 2];
_alertChunk = new float[framesPerTick * 2];
_mixBuffer = new float[framesPerTick * 2];
_delayedMix = new float[framesPerTick * 2];
_micChunk = new float[maxChunkFrames];
_loopbackChunk = new float[maxChunkFrames * 2];
_alertChunk = new float[maxChunkFrames * 2];
_mixBuffer = new float[maxChunkFrames * 2];
_delayedMix = new float[maxChunkFrames * 2];
_mic.Started += OnMicStarted;
_mic.SampleReady += OnMicSample;
@@ -393,12 +397,32 @@ public sealed class AudioMixer : IDisposable
var mixBuffer = _mixBuffer!;
var micChunk = _micChunk!;
var loopbackChunk = _loopbackChunk!;
// The audio clock is the monotonic clock, not the tick count. A fixed chunk per iteration turns
// any scheduler error into a permanent A/V rate error (measured 2026-11-12: 2.1% fast, 224 ms
// of drift in ten seconds); asking the clock how much is owed and producing exactly that keeps
// the track the same length as the show.
var clock = Stopwatch.StartNew();
long framesProduced = 0;
var maxChunkFrames = micChunk.Length;
while (!cancellationToken.IsCancellationRequested)
{
var nextTick = DateTime.UtcNow + _mixInterval;
try
{
var (_, micDrained, loopDrained) = FillAndMix(micChunk, loopbackChunk, mixBuffer);
var frames = AudioClock.FramesDue(
clock.Elapsed.TotalSeconds, framesProduced, OutputSampleRate, maxChunkFrames);
if (frames <= 0)
{
// Nothing owed yet — sooner than the tick interval, so a short yield, not a spin.
await Task.Delay(1, cancellationToken).ConfigureAwait(false);
continue;
}
var (_, micDrained, loopDrained) = FillAndMix(micChunk, loopbackChunk, mixBuffer, frames);
framesProduced += frames;
var samples = frames * 2;
// Negative offset: skip the armed |N| ms of stream head. The pipe
// writer tolerates a partial chunk (any float length is valid);
@@ -407,17 +431,18 @@ public sealed class AudioMixer : IDisposable
var writeFrom = 0;
if (_advanceSamplesRemaining > 0)
{
var drop = Math.Min(_advanceSamplesRemaining, mixBuffer.Length);
var drop = Math.Min(_advanceSamplesRemaining, samples);
_advanceSamplesRemaining -= drop;
writeFrom = drop;
}
if (writeFrom < mixBuffer.Length)
await pipe.WriteAsync(mixBuffer.AsMemory(writeFrom), cancellationToken).ConfigureAwait(false);
if (writeFrom < samples)
await pipe.WriteAsync(mixBuffer.AsMemory(writeFrom, samples - writeFrom),
cancellationToken).ConfigureAwait(false);
_micDrainedTotal += micDrained;
_loopDrainedTotal += loopDrained;
for (var i = writeFrom; i < mixBuffer.Length; i++)
for (var i = writeFrom; i < samples; i++)
{
var abs = Math.Abs(mixBuffer[i]);
if (abs > _peakMix) _peakMix = abs;
@@ -475,17 +500,18 @@ public sealed class AudioMixer : IDisposable
/// Returns the pre-duck mic RMS and how many samples each buffer contributed
/// (the 5s live-loop telemetry reads those to name whether capture or the
/// pipe starved a silent recording).</summary>
private (float MicRms, int MicDrained, int LoopDrained) FillAndMix(float[] micChunk, float[] loopbackChunk, float[] mix)
private (float MicRms, int MicDrained, int LoopDrained) FillAndMix(
float[] micChunk, float[] loopbackChunk, float[] mix, int frames)
{
var micCount = _micBuffer.Read(micChunk, micChunk.Length);
var loopCount = _loopbackBuffer.Read(loopbackChunk, loopbackChunk.Length);
var micCount = _micBuffer.Read(micChunk, frames);
var loopCount = _loopbackBuffer.Read(loopbackChunk, frames * 2);
var alertChunk = _alertChunk!;
var alertCount = _alertBuffer.Read(alertChunk, alertChunk.Length);
var alertCount = _alertBuffer.Read(alertChunk, frames * 2);
float micRms = 0;
if (micCount == 0)
{
Array.Clear(micChunk, 0, micChunk.Length);
Array.Clear(micChunk, 0, frames);
}
else
{
@@ -493,26 +519,26 @@ public sealed class AudioMixer : IDisposable
for (var i = 0; i < micCount; i++)
sumSquares += micChunk[i] * micChunk[i];
micRms = (float)Math.Sqrt(sumSquares / micCount);
if (micCount < micChunk.Length)
Array.Clear(micChunk, micCount, micChunk.Length - micCount);
if (micCount < frames)
Array.Clear(micChunk, micCount, frames - micCount);
}
if (loopCount == 0)
{
Array.Clear(loopbackChunk, 0, loopbackChunk.Length);
Array.Clear(loopbackChunk, 0, frames * 2);
}
else if (loopCount < loopbackChunk.Length)
else if (loopCount < frames * 2)
{
Array.Clear(loopbackChunk, loopCount, loopbackChunk.Length - loopCount);
Array.Clear(loopbackChunk, loopCount, frames * 2 - loopCount);
}
if (alertCount == 0)
{
Array.Clear(alertChunk, 0, alertChunk.Length);
Array.Clear(alertChunk, 0, frames * 2);
}
else if (alertCount < alertChunk.Length)
else if (alertCount < frames * 2)
{
Array.Clear(alertChunk, alertCount, alertChunk.Length - alertCount);
Array.Clear(alertChunk, alertCount, frames * 2 - alertCount);
}
// Mic-only delay: applied to the source, before the mix, so the desktop and alert audio keep
@@ -523,40 +549,40 @@ public sealed class AudioMixer : IDisposable
{
_micSyncDelay.Configure(micDelayMs);
var delayedMic = _delayedMic;
if (delayedMic is null || delayedMic.Length < micChunk.Length)
delayedMic = _delayedMic = new float[micChunk.Length];
_micSyncDelay.Process(micChunk, delayedMic);
Array.Copy(delayedMic, micChunk, micChunk.Length);
if (delayedMic is null || delayedMic.Length < frames)
delayedMic = _delayedMic = new float[frames];
_micSyncDelay.Process(micChunk.AsSpan(0, frames), delayedMic.AsSpan(0, frames));
Array.Copy(delayedMic, micChunk, frames);
}
var duck = _ducker.Update(micRms);
var micGain = (float)(_micGain?.Invoke() ?? 1.0);
var loopGain = (float)(_loopbackGain?.Invoke() ?? 1.0) * duck;
var frames = micChunk.Length;
for (var i = 0; i < frames; i++)
{
var m = micChunk[i] * micGain;
mix[i * 2] = m + loopbackChunk[i * 2] * loopGain + alertChunk[i * 2];
mix[i * 2 + 1] = m + loopbackChunk[i * 2 + 1] * loopGain + alertChunk[i * 2 + 1];
}
_masterLimiter.Process(mix);
var samples = frames * 2;
_masterLimiter.Process(mix.AsSpan(0, samples));
_syncDelay.Configure(_syncOffsetMs?.Invoke() ?? 0);
var delayed = _delayedMix;
if (delayed is null || delayed.Length < mix.Length)
delayed = _delayedMix = new float[mix.Length];
_syncDelay.Process(mix, delayed);
Array.Copy(delayed, mix, mix.Length);
if (delayed is null || delayed.Length < samples)
delayed = _delayedMix = new float[samples];
_syncDelay.Process(mix.AsSpan(0, samples), delayed.AsSpan(0, samples));
Array.Copy(delayed, mix, samples);
if (MonoOutput) Helpers.AudioFold.ToMono(mix.AsSpan(0, mix.Length), 2);
MeasureMaster(mix);
if (MonoOutput) Helpers.AudioFold.ToMono(mix.AsSpan(0, samples), 2);
MeasureMaster(mix.AsSpan(0, samples));
return (micRms, micCount, loopCount);
}
/// <summary>One pass over the finished mix: 1024 samples per channel, so the cost is noise
/// beside the mixer it sits behind.</summary>
private void MeasureMaster(float[] mix)
private void MeasureMaster(ReadOnlySpan<float> mix)
{
double left = 0, right = 0;
var frames = mix.Length / 2;
Services/Audio/MasterLimiter.cs
+1 -1
View File
@@ -21,7 +21,7 @@ public sealed class MasterLimiter
/// <summary>Scales the frame by the current gain. Gain drops instantly to
/// <c>ceiling/peak</c> when the frame exceeds the ceiling and recovers
/// toward unity slowly.</summary>
public void Process(float[] samples)
public void Process(Span<float> samples)
{
var peak = 0f;
foreach (var s in samples)