Merge pull request #7 from TensorStack-AI/SupertonicTTS

Supertonic TTS

Author: saddam213

TensorStack.Audio.Windows/AudioInput.cs

@@ -12,7 +12,7 @@ namespace TensorStack.Audio
     /// </summary>
     public class AudioInput : AudioInputBase
     {
-        private readonly string _sourceFile;
+        private string _sourceFile;
 
         /// <summary>
         /// Initializes a new instance of the <see cref="AudioInput"/> class.
@@ -21,6 +21,13 @@ public class AudioInput : AudioInputBase
         public AudioInput(string filename, string audioCodec = "pcm_s16le", int sampleRate = 16000, int channels = 1)
             : this(filename, AudioManager.LoadTensor(filename, audioCodec, sampleRate, channels)) { }
 
+        /// <summary>
+        /// Initializes a new instance of the <see cref="AudioInput"/> class.
+        /// </summary>
+        /// <param name="audioTensor">The audio tensor.</param>
+        public AudioInput(AudioTensor audioTensor)
+            : base(audioTensor) { }
+
         /// <summary>
         /// Initializes a new instance of the <see cref="AudioInput"/> class.
         /// </summary>
@@ -44,6 +51,9 @@ protected AudioInput(string filename, AudioTensor audioTensor)
         /// <param name="filename">The filename.</param>
         public override void Save(string filename)
         {
+            if (string.IsNullOrEmpty(_sourceFile))
+                _sourceFile = filename;
+
             AudioManager.SaveAudio(filename, this);
         }
 
@@ -55,6 +65,9 @@ public override void Save(string filename)
         /// <param name="cancellationToken">The cancellation token.</param>
         public override async Task SaveAsync(string filename, CancellationToken cancellationToken = default)
         {
+            if (string.IsNullOrEmpty(_sourceFile))
+                _sourceFile = filename;
+
             await AudioManager.SaveAudioAync(filename, this, cancellationToken);
         }
 

TensorStack.TextGeneration/Pipelines/Supertonic/README.md

@@ -0,0 +1,19 @@
+# Supertonic TTS
+https://github.com/supertone-inc/supertonic
+
+
+```csharp
+// [model] https://huggingface.co/TensorStack/Supertonic-onnx
+
+var provider = Provider.GetProvider(GraphOptimizationLevel.ORT_ENABLE_ALL);
+var modelPath = "M:\\Models\\Supertonic-onnx";
+var pipeline = SupertonicPipeline.Create(modelPath, provider);
+var options = new SupertonicOptions
+{
+    TextInput = "On a quiet morning in the old town, a clockmaker named Ellis unlocked his tiny shop",
+    VoiceStyle = "Female1"
+};
+
+var generateResult = await pipeline.RunAsync(options);
+AudioManager.SaveAudio("Output.wav", generateResult);
+```

TensorStack.TextGeneration/Pipelines/Supertonic/SupertonicConfig.cs

@@ -0,0 +1,20 @@
+using TensorStack.Common;
+
+namespace TensorStack.TextGeneration.Pipelines.Supertonic
+{
+    public record SupertonicConfig
+    {
+        public int SampleRate { get; init; } = 44100;
+        public int BaseChunkSize { get; init; } = 512;
+        public int LatentDim { get; init; } = 24;
+        public int ChunkCompressFactor { get; init; } = 6;
+        public int TextEmbedSize { get; init; } = 256;
+        public int ScaleFactor { get; init; } = 3072;
+        public string IndexerPath { get; init; }
+        public string VoiceStylePath { get; init; }
+        public ModelConfig PredictorConfig { get; init; }
+        public ModelConfig EncoderConfig { get; init; }
+        public ModelConfig EstimatorConfig { get; init; }
+        public ModelConfig DecoderConfig { get; init; }
+    }
+}

TensorStack.TextGeneration/Pipelines/Supertonic/SupertonicOptions.cs

@@ -0,0 +1,13 @@
+using TensorStack.Common.Pipeline;
+
+namespace TensorStack.TextGeneration.Pipelines.Supertonic
+{
+    public record SupertonicOptions : IRunOptions
+    {
+        public string TextInput { get; set; }
+        public string VoiceStyle { get; set; }
+        public int Steps { get; set; } = 5;
+        public float Speed { get; set; } = 1f;
+        public float SilenceDuration { get; set; } = 0.3f;
+    }
+}