Merge pull request #20 from oxygen-dioxide/diffsinger-enunux

Diffsinger enunux

Author: oxygen-dioxide

OpenUtau.Core/Enunu/EnunuConfig.cs

@@ -56,15 +56,17 @@ public EnunuConfig Convert() {
             enunuConfig.sampleRate = this.sampleRate;
             enunuConfig.framePeriod = this.framePeriod;
             enunuConfig.extensions = new EnunuExtensions();
-            ParseEnunuExtension(enunuConfig.extensions.ust_editor, this.extensions.ust_editor);
-            ParseEnunuExtension(enunuConfig.extensions.ust_converter, this.extensions.ust_converter);
-            ParseEnunuExtension(enunuConfig.extensions.score_editor, this.extensions.score_editor);
-            ParseEnunuExtension(enunuConfig.extensions.timing_calculator, this.extensions.timing_calculator);
-            ParseEnunuExtension(enunuConfig.extensions.timing_editor, this.extensions.timing_editor);
-            ParseEnunuExtension(enunuConfig.extensions.acoustic_calculator, this.extensions.acoustic_calculator);
-            ParseEnunuExtension(enunuConfig.extensions.acoustic_editor, this.extensions.acoustic_editor);
-            ParseEnunuExtension(enunuConfig.extensions.wav_synthesizer, this.extensions.wav_synthesizer);
-            ParseEnunuExtension(enunuConfig.extensions.wav_editor, this.extensions.wav_editor);
+            if (this.extensions != null) {
+                ParseEnunuExtension(enunuConfig.extensions.ust_editor, this.extensions.ust_editor);
+                ParseEnunuExtension(enunuConfig.extensions.ust_converter, this.extensions.ust_converter);
+                ParseEnunuExtension(enunuConfig.extensions.score_editor, this.extensions.score_editor);
+                ParseEnunuExtension(enunuConfig.extensions.timing_calculator, this.extensions.timing_calculator);
+                ParseEnunuExtension(enunuConfig.extensions.timing_editor, this.extensions.timing_editor);
+                ParseEnunuExtension(enunuConfig.extensions.acoustic_calculator, this.extensions.acoustic_calculator);
+                ParseEnunuExtension(enunuConfig.extensions.acoustic_editor, this.extensions.acoustic_editor);
+                ParseEnunuExtension(enunuConfig.extensions.wav_synthesizer, this.extensions.wav_synthesizer);
+                ParseEnunuExtension(enunuConfig.extensions.wav_editor, this.extensions.wav_editor);
+            }
             return enunuConfig;
         }
 

OpenUtau.Plugin.Builtin/EnunuOnnx/EnunuConfig.cs

@@ -0,0 +1,59 @@
+using System.Collections.Generic;
+using System.IO;
+using OpenUtau.Core;
+
+//Instead of using Enunu/EnunuConfig.cs, we created a copy to add EnunuOnnx-specific features
+//without potentially breaking the existing Enunu Phonemizers and renderers.
+namespace OpenUtau.Plugin.Builtin.EnunuOnnx {
+    class EnunuConfig {
+        public string tablePath;
+        public string questionPath;
+        public int sampleRate;
+        public double framePeriod;
+        public string modelDir;
+        public string statsDir;
+        public EnunuDurationConfig duration;
+        public EnunuTimelagConfig timelag;
+
+        public static EnunuConfig Load(string configPath) {
+            var configTxt = File.ReadAllText(configPath);
+            RawEnunuConfig config = Yaml.DefaultDeserializer.Deserialize<RawEnunuConfig>(configTxt);
+            return config.Convert();
+        }
+    }
+
+
+    class EnunuTimelagConfig {
+        public string checkpoint;
+        public List<int> allowedRange;
+        public List<int> allowedRangeRest;
+    }
+    class EnunuDurationConfig {
+        public string checkpoint;
+
+    }
+
+    class RawEnunuConfig {
+        public string tablePath;
+        public string questionPath;
+        public int sampleRate;
+        public double framePeriod;
+        public string modelDir;
+        public string statsDir;
+        public EnunuDurationConfig duration;
+        public EnunuTimelagConfig timelag;
+
+        public EnunuConfig Convert() {
+            EnunuConfig enunuConfig = new EnunuConfig();
+            enunuConfig.tablePath = this.tablePath;
+            enunuConfig.questionPath = this.questionPath;
+            enunuConfig.sampleRate = this.sampleRate;
+            enunuConfig.framePeriod = this.framePeriod;
+            enunuConfig.duration = this.duration;
+            enunuConfig.modelDir = this.modelDir;
+            enunuConfig.statsDir = this.statsDir;
+            enunuConfig.timelag = this.timelag;
+            return enunuConfig;
+        }
+    }
+}

OpenUtau.Plugin.Builtin/EnunuOnnx/EnunuOnnxEnglishPhonemizer.cs

@@ -0,0 +1,29 @@
+using System;
+using System.Collections.Generic;
+using System.IO;
+using System.Text;
+using OpenUtau.Api;
+using OpenUtau.Core.G2p;
+using Serilog;
+
+namespace OpenUtau.Plugin.Builtin {
+    [Phonemizer("Enunu Onnx English Phonemizer", "ENUNU X EN", language:"EN")]
+    public class EnunuOnnxEnglishPhonemizer : EnunuOnnxPhonemizer {
+        protected override IG2p LoadG2p(string rootPath) {
+            var g2ps = new List<IG2p>();
+
+            // Load dictionary from singer folder.
+            string file = Path.Combine(rootPath, "enunux.yaml");
+            if (File.Exists(file)) {
+                try {
+                    g2ps.Add(G2pDictionary.NewBuilder().Load(File.ReadAllText(file)).Build());
+                } catch (Exception e) {
+                    Log.Error(e, $"Failed to load {file}");
+                }
+            }
+            g2ps.Add(new ArpabetG2p());
+            return new G2pFallbacks(g2ps.ToArray());
+        }
+
+    }
+}