Improve meeting audio mixing pipeline

2026-05-29 13:47:21 +02:00
parent 9d0f6a1295
commit 740f93f185
18 changed files with 870 additions and 136 deletions
@@ -28,74 +28,74 @@ public sealed class AzureSpeechStreamingTranscriptionProvider : IStreamingTransc
        var enumerator = audio.GetAsyncEnumerator(cancellationToken);
        try
        {
-        if (!await enumerator.MoveNextAsync())
-        {
-            yield break;
-        }
-
-        var firstChunk = enumerator.Current;
-        using var pushStream = AudioInputStream.CreatePushStream(
-            AudioStreamFormat.GetWaveFormatPCM(
-                (uint)firstChunk.SampleRate,
-                16,
-                (byte)firstChunk.Channels));
-        using var audioConfig = AudioConfig.FromStreamInput(pushStream);
-        var speechConfig = CreateSpeechConfig();
-        using var recognizer = CreateTranscriber(speechConfig, audioConfig);
-        AddPhraseList(recognizer, pipelineOptions.DictationWords);
-        var segments = Channel.CreateUnbounded<TranscriptionSegment>();
-
-        recognizer.Transcribed += (_, args) =>
-        {
-            if (args.Result.Reason != ResultReason.RecognizedSpeech
-                || string.IsNullOrWhiteSpace(args.Result.Text))
+            if (!await enumerator.MoveNextAsync())
            {
-                return;
+                yield break;
            }

-            var start = TimeSpan.FromTicks(args.Result.OffsetInTicks);
-            var segment = new TranscriptionSegment(
-                start,
-                start + args.Result.Duration,
-                NormalizeSpeakerId(args.Result.SpeakerId),
-                args.Result.Text.Trim());
-            logger.LogInformation(
-                "Azure Speech emitted segment at {Start} from {Speaker}: {Text}",
-                segment.Start,
-                segment.Speaker,
-                segment.Text);
-            segments.Writer.TryWrite(segment);
-        };
-        recognizer.Canceled += (_, args) =>
-        {
-            if (args.Reason == CancellationReason.Error)
+            var firstChunk = enumerator.Current;
+            using var pushStream = AudioInputStream.CreatePushStream(
+                AudioStreamFormat.GetWaveFormatPCM(
+                    (uint)firstChunk.SampleRate,
+                    16,
+                    (byte)firstChunk.Channels));
+            using var audioConfig = AudioConfig.FromStreamInput(pushStream);
+            var speechConfig = CreateSpeechConfig();
+            using var recognizer = CreateTranscriber(speechConfig, audioConfig);
+            AddPhraseList(recognizer, pipelineOptions.DictationWords);
+            var segments = Channel.CreateUnbounded<TranscriptionSegment>();
+
+            recognizer.Transcribed += (_, args) =>
            {
-                segments.Writer.TryComplete(new InvalidOperationException(
-                    $"Azure Speech recognition canceled: {args.ErrorCode} {args.ErrorDetails}"));
-                return;
+                if (args.Result.Reason != ResultReason.RecognizedSpeech
+                    || string.IsNullOrWhiteSpace(args.Result.Text))
+                {
+                    return;
+                }
+
+                var start = TimeSpan.FromTicks(args.Result.OffsetInTicks);
+                var segment = new TranscriptionSegment(
+                    start,
+                    start + args.Result.Duration,
+                    NormalizeSpeakerId(args.Result.SpeakerId),
+                    args.Result.Text.Trim());
+                logger.LogInformation(
+                    "Azure Speech emitted segment at {Start} from {Speaker}: {Text}",
+                    segment.Start,
+                    segment.Speaker,
+                    segment.Text);
+                segments.Writer.TryWrite(segment);
+            };
+            recognizer.Canceled += (_, args) =>
+            {
+                if (args.Reason == CancellationReason.Error)
+                {
+                    segments.Writer.TryComplete(new InvalidOperationException(
+                        $"Azure Speech recognition canceled: {args.ErrorCode} {args.ErrorDetails}"));
+                    return;
+                }
+
+                segments.Writer.TryComplete();
+            };
+
+            await recognizer.StartTranscribingAsync();
+            var pumpTask = Task.Run(
+                () => PumpAudioAsync(
+                    firstChunk,
+                    enumerator,
+                    pushStream,
+                    recognizer,
+                    segments.Writer,
+                    options.AzureSpeech.RecognitionStopTimeout,
+                    cancellationToken),
+                CancellationToken.None);
+
+            await foreach (var segment in segments.Reader.ReadAllAsync(cancellationToken))
+            {
+                yield return segment;
            }

-            segments.Writer.TryComplete();
-        };
-
-        await recognizer.StartTranscribingAsync();
-        var pumpTask = Task.Run(
-            () => PumpAudioAsync(
-                firstChunk,
-                enumerator,
-                pushStream,
-                recognizer,
-                segments.Writer,
-                options.AzureSpeech.RecognitionStopTimeout,
-                cancellationToken),
-            CancellationToken.None);
-
-        await foreach (var segment in segments.Reader.ReadAllAsync(cancellationToken))
-        {
-            yield return segment;
-        }
-
-        await pumpTask.WaitAsync(cancellationToken);
+            await pumpTask.WaitAsync(cancellationToken);
        }
        finally
        {