RE1-T115 TTS fixes

Tests/Resgrid.Tests/Web/Tts/TtsServiceTests.cs

@@ -253,11 +253,11 @@ public void create_espeak_start_info_should_use_mbrola_profile_for_english_voice
 				"-v",
 				"mb-us1",
 				"-s",
-				"130",
+				"140",
 				"-p",
 				"50",
 				"-g",
-				"3");
+				"2");
 		}
 
 		[Test]
@@ -308,7 +308,8 @@ private static AudioProcessingService CreateService()
 		{
 			return new AudioProcessingService(
 				Options.Create(new TtsOptions()),
-				Mock.Of<ILogger<AudioProcessingService>>());
+				Mock.Of<ILogger<AudioProcessingService>>(),
+				Mock.Of<ITextPreprocessor>());
 		}
 
 		private static T InvokePrivateMethod<T>(object instance, string methodName, params object[] arguments)

Web/Resgrid.Web.Tts/Program.cs

@@ -1,6 +1,3 @@
-using Amazon;
-using Amazon.Runtime;
-using Amazon.S3;
 using Microsoft.AspNetCore.Diagnostics.HealthChecks;
 using Microsoft.AspNetCore.HttpOverrides;
 using Microsoft.AspNetCore.Mvc;
@@ -106,12 +103,8 @@ await context.HttpContext.Response.WriteAsJsonAsync(
 	};
 });
 
-builder.Services.AddSingleton<IAmazonS3>(sp =>
-{
-	var options = sp.GetRequiredService<IOptions<S3StorageOptions>>().Value;
-	return CreateS3Client(options);
-});
 builder.Services.AddSingleton<IStorageService, S3StorageService>();
+builder.Services.AddSingleton<ITextPreprocessor, TextPreprocessor>();
 builder.Services.AddSingleton<ICacheService, CacheService>();
 builder.Services.AddSingleton<IAudioProcessingService, AudioProcessingService>();
 builder.Services.AddSingleton<ITtsPlaybackUrlService, TtsPlaybackUrlService>();
@@ -156,34 +149,4 @@ await context.HttpContext.Response.WriteAsJsonAsync(
 
 app.Run();
 
-static AmazonS3Client CreateS3Client(S3StorageOptions options)
-{
-	var credentials = new BasicAWSCredentials(options.AccessKey, options.SecretKey);
-	var config = new AmazonS3Config
-	{
-		ForcePathStyle = options.ForcePathStyle,
-		AuthenticationRegion = options.Region
-	};
-
-	if (!string.IsNullOrWhiteSpace(options.Endpoint))
-	{
-		if (Uri.TryCreate(options.Endpoint, UriKind.Absolute, out var endpointUri))
-		{
-			config.ServiceURL = endpointUri.GetLeftPart(UriPartial.Authority);
-			config.UseHttp = endpointUri.Scheme.Equals(Uri.UriSchemeHttp, StringComparison.OrdinalIgnoreCase);
-		}
-		else
-		{
-			config.ServiceURL = $"{(options.UseSsl ? Uri.UriSchemeHttps : Uri.UriSchemeHttp)}://{options.Endpoint}";
-			config.UseHttp = !options.UseSsl;
-		}
-	}
-	else
-	{
-		config.RegionEndpoint = RegionEndpoint.GetBySystemName(options.Region);
-	}
-
-	return new AmazonS3Client(credentials, config);
-}
-
 public partial class Program;

Web/Resgrid.Web.Tts/Resgrid.Web.Tts.csproj

@@ -10,7 +10,6 @@
 	</PropertyGroup>
 
 	<ItemGroup>
-		<PackageReference Include="AWSSDK.S3" Version="3.7.414.5" />
 		<PackageReference Include="Microsoft.Extensions.Caching.StackExchangeRedis" Version="9.0.2" />
 		<PackageReference Include="Sentry.AspNetCore" Version="5.4.0" />
 		<PackageReference Include="Sentry.Profiling" Version="5.4.0" />

Web/Resgrid.Web.Tts/Services/AudioProcessingService.cs

@@ -9,20 +9,23 @@ namespace Resgrid.Web.Tts.Services
 	public sealed class AudioProcessingService : IAudioProcessingService
 	{
 		private const string MbrolaEnglishVoice = "mb-us1";
-		private const int MbrolaEnglishSpeed = 130;
+		private const int MbrolaEnglishSpeed = 140;
 		private const int MbrolaEnglishPitch = 50;
-		private const int MbrolaEnglishWordGap = 3;
+		private const int MbrolaEnglishWordGap = 2;
 		private const string TelephoneAudioFilter = "highpass=f=200, lowpass=f=3000, anequalizer=c0 f=2500 w=1000 g=3 t=1";
 
 		private readonly TtsOptions _options;
 		private readonly ILogger<AudioProcessingService> _logger;
+		private readonly ITextPreprocessor _textPreprocessor;
 
 		public AudioProcessingService(
 			IOptions<TtsOptions> options,
-			ILogger<AudioProcessingService> logger)
+			ILogger<AudioProcessingService> logger,
+			ITextPreprocessor textPreprocessor)
 		{
 			_options = options.Value;
 			_logger = logger;
+			_textPreprocessor = textPreprocessor;
 		}
 
 		public async Task<byte[]> GenerateNormalizedWavAsync(string text, string voice, int speed, CancellationToken cancellationToken)
@@ -36,7 +39,8 @@ public async Task<byte[]> GenerateNormalizedWavAsync(string text, string voice,
 
 			try
 			{
-				await RunEspeakAsync(text, voice, speed, rawFilePath, cancellationToken);
+				var preprocessedText = _textPreprocessor.Preprocess(text, voice);
+				await RunEspeakAsync(preprocessedText, voice, speed, rawFilePath, cancellationToken);
 				await RunFfmpegAsync(rawFilePath, normalizedFilePath, cancellationToken);
 
 				return await File.ReadAllBytesAsync(normalizedFilePath, cancellationToken);

Web/Resgrid.Web.Tts/Services/ITextPreprocessor.cs

@@ -0,0 +1,15 @@
+namespace Resgrid.Web.Tts.Services
+{
+	/// <summary>
+	/// Preprocesses dispatch text before it is sent to the TTS engine so that
+	/// common abbreviations, codes, and jargon are spoken intelligibly.
+	/// </summary>
+	public interface ITextPreprocessor
+	{
+		/// <summary>
+		/// Normalises the input text for the given voice / language so that eSpeak
+		/// (or any downstream TTS engine) produces the most natural speech.
+		/// </summary>
+		string Preprocess(string text, string voice);
+	}
+}