Merge pull request #10 from oxygen-dioxide/diffsinger-nomidi-directml

tailMs and consonant duration limit
xunmengshe · Jan 17, 2023 · f7ba9a4 · f7ba9a4
2 parents f4f458c + a0a7c18
commit f7ba9a4
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 5 deletions.
diff --git a/OpenUtau.Core/DiffSinger/DiffSingerMandarinPhonemizer.cs b/OpenUtau.Core/DiffSinger/DiffSingerMandarinPhonemizer.cs
@@ -71,10 +71,14 @@ public override Result Process(Note[] notes, Note? prev, Note? next, Note? prevN
             //使用vogen的辅音时间
             Result VogenResult = base.Process(notes, prev, next, prevNeighbour, nextNeighbour, prevs);
             //辅音长度至少为1帧
+            int consonantPos = Math.Min(VogenResult.phonemes[0].position, -frameTick);
+            //辅音长度不能超过上一音符时长的2/3
+            if (prevNeighbour != null) {
+                consonantPos = Math.Max(consonantPos, -prevNeighbour.Value.duration * 2 / 3);
+            }
             return new Result {
                 phonemes = new Phoneme[] {
-                    new Phoneme {phoneme = phones.Item1, 
-                        position = Math.Min(VogenResult.phonemes[0].position,-frameTick)},
+                    new Phoneme {phoneme = phones.Item1, position = consonantPos},
                     new Phoneme {phoneme = phones.Item2, position = 0}
                 },
             };

diff --git a/OpenUtau.Core/DiffSinger/DiffSingerRenderer.cs b/OpenUtau.Core/DiffSinger/DiffSingerRenderer.cs
@@ -16,7 +16,7 @@
 namespace OpenUtau.Core.DiffSinger {
     public class DiffSingerRenderer : IRenderer {
         const float headMs = 0;
-        const float tailMs = 0;
+        const float tailMs = 500;
 
         static readonly HashSet<string> supportedExp = new HashSet<string>(){
             Format.Ustx.DYN,
@@ -109,10 +109,12 @@ float[] InvokeDiffsinger(RenderPhrase phrase,int speedup) {
             //speedup：加速倍数
             var tokens = phrase.phones
                 .Select(p => p.phoneme)
+                .Append("SP")
                 .Select(x => (long)(singer.phonemes.IndexOf(x)))
                 .ToList();
             var durations = phrase.phones
                 .Select(p => (long)(p.endMs / frameMs) - (long)(p.positionMs / frameMs))//防止累计误差
+                .Append(tailFrames)
                 .ToList();
             var totalFrames = (int)(durations.Sum());
             var f0 = SampleCurve(phrase, phrase.pitches, 0, totalFrames, headFrames, tailFrames, x => MusicMath.ToneToFreq(x * 0.01));
@@ -167,8 +169,8 @@ double[] SampleCurve(RenderPhrase phrase, float[] curve, double defaultValue, in
                 }
             }
             //填充头尾
-            Array.Fill(result, defaultValue, 0, headFrames);
-            Array.Fill(result, defaultValue, length - tailFrames, tailFrames);
+            Array.Fill(result, convert(curve[0]), 0, headFrames);
+            Array.Fill(result, convert(curve.Last()), length - tailFrames, tailFrames);
             return result;
         }