Merge remote-tracking branch 'upstream/master'

atonyxu · Feb 29, 2024 · a3a7511 · a3a7511
2 parents 0c5a16e + c49715e
commit a3a7511
Show file tree

Hide file tree

Showing 56 changed files with 1,289 additions and 870 deletions.
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -69,7 +69,7 @@ jobs:
     - name: download vocoder
       uses: carlosperate/download-file-action@v1.0.3
       with:
-        file-url: https://github.com/xunmengshe/OpenUtau/releases/download/0.0.0.0/nsf_hifigan.oudep
+        file-url: https://github.com/xunmengshe/OpenUtau/releases/download/0.0.0.0/nsf_hifigan-2024.02.oudep
         file-name: nsf_hifigan.oudep
         location: ./
 

diff --git a/OpenUtau.Core/Api/Phonemizer.cs b/OpenUtau.Core/Api/Phonemizer.cs
@@ -120,6 +120,8 @@ public struct Phoneme {
             /// </summary>
             public PhonemeAttributes attributes;
 
+            public int? index;
+
             public override string ToString() => $"\"{phoneme}\" pos:{position}";
         }
 

diff --git a/OpenUtau.Core/Classic/ClassicSinger.cs b/OpenUtau.Core/Classic/ClassicSinger.cs
@@ -77,7 +77,7 @@ public override void Reload() {
         void Load() {
             if (Avatar != null && File.Exists(Avatar)) {
                 try {
-                    using (var stream = new FileStream(Avatar, FileMode.Open)) {
+                    using (var stream = new FileStream(Avatar, FileMode.Open, FileAccess.Read)) {
                         using (var memoryStream = new MemoryStream()) {
                             stream.CopyTo(memoryStream);
                             avatarData = memoryStream.ToArray();

diff --git a/OpenUtau.Core/Classic/VoicebankErrorChecker.cs b/OpenUtau.Core/Classic/VoicebankErrorChecker.cs
@@ -101,20 +101,15 @@ public void Check() {
                     message = $"There are duplicate aliases.{message}"
                 });
             }
-            //Cross platform check
-            //Windows path is case insensitive, while MacOS path and Linux path are case sensitive.
-            //On Windows, check if the wave filename in oto.ini is the same as the filename in the file system.
-            if(OS.IsWindows()){
-                foreach(var otoSet in voicebank.OtoSets) {
-                    WindowsCaseCheck(otoSet);
-                }
-                WindowsCaseCheck(voicebank.BasePath, new string[]{
-                    "chatacter.txt", 
-                    "character.yaml", 
-                    "prefix.map",
-                    });
+            foreach(var otoSet in voicebank.OtoSets) {
+                CheckCaseMatchForFileReference(otoSet);
+                CheckDuplicatedNameIgnoringCase(otoSet);
             }
-            //TODO: On MacOS and Linux, check if there are files that have the same name but different case.
+            CheckCaseMatchForFileReference(voicebank.BasePath, new string[]{
+                "chatacter.txt", 
+                "character.yaml", 
+                "prefix.map",
+                });
         }
 
         bool TryGetFileDuration(string filePath, Oto oto, out double fileDuration) {
@@ -260,16 +255,16 @@ bool FindDuplication(out List<Oto> duplicates) {
         /// </summary>
         /// <param name="otoSet">otoSet to be checked</param>
         /// <returns></returns>
-        bool WindowsCaseCheck(OtoSet otoSet) {
-            return WindowsCaseCheck(
+        bool CheckCaseMatchForFileReference(OtoSet otoSet) {
+            return CheckCaseMatchForFileReference(
                 Directory.GetParent(otoSet.File).FullName, 
                 otoSet.Otos
                     .Select(oto => oto.Wav)
                     .Append(otoSet.File)//oto.ini itself
                     .ToHashSet());
         }
 
-        bool WindowsCaseCheck(string folder, IEnumerable<string> correctFileNames){
+        bool CheckCaseMatchForFileReference(string folder, IEnumerable<string> correctFileNames){
             bool valid = true;
             Dictionary<string, string> fileNamesLowerToActual = Directory.GetFiles(folder)
                 .Select(Path.GetFileName)
@@ -280,15 +275,36 @@ bool WindowsCaseCheck(string folder, IEnumerable<string> correctFileNames){
                 }
                 if (fileNamesLowerToActual[fileName.ToLower()] != fileName) {
                     valid = false;
-                    Infos.Add(new VoicebankError() {
+                    Errors.Add(new VoicebankError() {
                         message = $"Wrong case in file name: \n"
                             + $"expected: {Path.Join(folder,fileName)}\n"
                             + $"Actual: {Path.Join(folder,fileNamesLowerToActual[fileName.ToLower()])}\n"
-                            + $"voicebank may not work on another OS."
+                            + $"The voicebank may not work on another OS."
                     });
                 }
             }
             return valid;
         }
+
+        /// <summary>
+        /// Check if the file names are duplicated when converted to lower case.
+        /// </summary>
+        /// <param name="otoSet">otoSet to be checked</param>
+        /// <returns></returns>
+        bool CheckDuplicatedNameIgnoringCase(OtoSet otoSet) {
+            var wavNames = otoSet.Otos.Select(x => x.Wav).Distinct().ToList();
+            var duplicatedGroups = wavNames.GroupBy(x => x.ToLower())
+                .Where(group => group.Count() > 1)
+                .ToList();
+            foreach (var group in duplicatedGroups) {
+                Errors.Add(new VoicebankError() {
+                    message = $"Duplicated file names found when ignoreing case in oto set \"{otoSet.Name}\":"
+                    + string.Join(", ", group.Select(x => $"\"{x}\""))
+                    + ".\n"
+                    + "The voicebank may not work on another OS with case-sensitivity."
+                });
+            }
+            return duplicatedGroups.Count == 0;
+        }
     }
 }
diff --git a/OpenUtau.Core/Classic/VoicebankPublisher.cs b/OpenUtau.Core/Classic/VoicebankPublisher.cs
@@ -0,0 +1,83 @@
+using Ignore;
+using System;
+using System.Collections.Generic;
+using System.IO;
+using System.IO.Compression;
+using System.Linq;
+
+using OpenUtau.Core.Ustx;
+
+namespace OpenUtau.Classic {
+    public class VoicebankPublisher {
+        private readonly Action<double, string> progress;
+        private readonly Ignore.Ignore? ignore;
+
+        public VoicebankPublisher(Action<double, string> progress, string? gitIgnore) {
+            this.progress = progress;
+            if(gitIgnore != null) {
+                ignore = new Ignore.Ignore();
+                ignore.Add(gitIgnore.Split("\n"));
+            }
+        }
+
+        private static void ModifyConfig(USinger singer, Action<VoicebankConfig> modify) {
+            var yamlFile = Path.Combine(singer.Location, "character.yaml");
+            VoicebankConfig? config = null;
+            if (File.Exists(yamlFile)) {
+                using (var stream = File.OpenRead(yamlFile)) {
+                    config = VoicebankConfig.Load(stream);
+                }
+            }
+            if (config == null) {
+                config = new VoicebankConfig();
+            }
+            modify(config);
+            using (var stream = File.Open(yamlFile, FileMode.Create)) {
+                config.Save(stream);
+            }
+        }
+
+        private bool IsIgnored(string relativePath){
+            return ignore?.IsIgnored(relativePath.Replace('\\', '/')) ?? false;
+        }
+
+        private List<string> GetFilesToPack(string singerPath)
+        {
+            List<string> fileList = Directory.EnumerateFiles(singerPath, "*.*", SearchOption.AllDirectories).ToList();
+            List<string> packList = fileList.FindAll(x => !IsIgnored(System.IO.Path.GetRelativePath(singerPath, x)));
+            return packList;
+        }
+
+        ///<summary>
+        ///Compress a voicebank into an optimized zip archive for distribution.
+        ///This function only supports voicebanks that follow the classic packaging model,
+        ///including utau, enunu and diffsinger.
+        ///Vogen voicebanks aren't supported.
+        ///</summary>
+        public void Publish(USinger singer, string outputFile){
+            var location = singer.Location;
+            if(!Directory.Exists(location)){
+                return;
+            }
+            progress.Invoke(0, $"Publishing {singer.Name}");
+            //Write singer type into character.yaml
+            try {
+                ModifyConfig(singer, config => config.SingerType = singer.SingerType.ToString().ToLower());
+            } catch (Exception e) {  }
+            var packList = GetFilesToPack(location);
+            int index = 0;
+            int fileCount = packList.Count();
+            using(ZipArchive archive = new ZipArchive(File.Create(outputFile), ZipArchiveMode.Create))
+            {
+                foreach (var absFilePath in packList)
+                {
+                    index++;
+                    progress.Invoke(100.0 * index / fileCount, $"Compressing {absFilePath}");
+                    string reFilePath = Path.GetRelativePath(location, absFilePath);
+                    archive.CreateEntryFromFile(absFilePath, reFilePath);
+                }
+            }
+            progress.Invoke(0, $"Published {singer.Name} to {outputFile}");
+        }
+    }
+}
diff --git a/OpenUtau.Core/Commands/ExpCommands.cs b/OpenUtau.Core/Commands/ExpCommands.cs
@@ -1,7 +1,6 @@
 using System;
 using System.Collections.Generic;
 using System.Linq;
-
 using OpenUtau.Core.Ustx;
 using OpenUtau.Core.Util;
 
@@ -24,15 +23,15 @@ public ExpCommand(UVoicePart part) {
     public class SetNoteExpressionCommand : ExpCommand {
         public readonly UProject project;
         public readonly UTrack track;
-        public readonly float[] newValue;
-        public readonly float[] oldValue;
-        public SetNoteExpressionCommand(UProject project, UTrack track, UVoicePart part, UNote note, string abbr, float[] values) : base(part) {
+        public readonly float?[] newValue;
+        public readonly float?[] oldValue;
+        public SetNoteExpressionCommand(UProject project, UTrack track, UVoicePart part, UNote note, string abbr, float?[] values) : base(part) {
             this.project = project;
             this.track = track;
             this.Note = note;
             Key = abbr;
             newValue = values;
-            oldValue = note.GetExpression(project, track, abbr).Select(t => t.Item1).ToArray();
+            oldValue = note.GetExpressionNoteHas(project, track, abbr);
         }
         public override string ToString() => $"Set note expression {Key}";
         public override void Execute() => Note.SetExpression(project, track, Key, newValue);
@@ -47,21 +46,26 @@ public class SetPhonemeExpressionCommand : ExpCommand {
         public readonly UProject project;
         public readonly UTrack track;
         public readonly UPhoneme phoneme;
-        public readonly float newValue;
-        public readonly float oldValue;
+        public readonly float? newValue;
+        public readonly float? oldValue;
         public override ValidateOptions ValidateOptions
             => new ValidateOptions {
                 SkipTiming = true,
                 Part = Part,
                 SkipPhonemizer = !needsPhonemizer.Contains(Key),
             };
-        public SetPhonemeExpressionCommand(UProject project, UTrack track, UVoicePart part, UPhoneme phoneme, string abbr, float value) : base(part) {
+        public SetPhonemeExpressionCommand(UProject project, UTrack track, UVoicePart part, UPhoneme phoneme, string abbr, float? value) : base(part) {
             this.project = project;
             this.track = track;
             this.phoneme = phoneme;
             Key = abbr;
             newValue = value;
-            oldValue = phoneme.GetExpression(project, track, abbr).Item1;
+            var oldExp = phoneme.GetExpression(project, track, abbr);
+            if (oldExp.Item2) {
+                oldValue = oldExp.Item1;
+            } else {
+                oldValue = null;
+            }
         }
         public override string ToString() => $"Set phoneme expression {Key}";
         public override void Execute() {

diff --git a/OpenUtau.Core/DiffSinger/DiffSingerBasePhonemizer.cs b/OpenUtau.Core/DiffSinger/DiffSingerBasePhonemizer.cs
@@ -88,7 +88,7 @@ string[] GetSymbols(Note note) {
             //1. phonetic hint
             //2. query from g2p dictionary
             //3. treat lyric as phonetic hint, including single phoneme
-            //4. default pause
+            //4. empty
             if (!string.IsNullOrEmpty(note.phoneticHint)) {
                 // Split space-separated symbols into an array.
                 return note.phoneticHint.Split()
@@ -108,7 +108,7 @@ string[] GetSymbols(Note note) {
             if (lyricSplited.Length > 0) {
                 return lyricSplited;
             }
-            return new string[] { defaultPause };
+            return new string[] { };
         }
 
         string GetSpeakerAtIndex(Note note, int index){
@@ -122,21 +122,20 @@ string GetSpeakerAtIndex(Note note, int index){
             return speaker.Suffix;
         }
 
-        dsPhoneme[] GetDsPhonemes(Note note){
-            return GetSymbols(note)
-                .Select((symbol, index) => new dsPhoneme(symbol, GetSpeakerAtIndex(note, index)))
-                .ToArray();            
-        }
-
         protected bool IsSyllableVowelExtensionNote(Note note) {
             return note.lyric.StartsWith("+~") || note.lyric.StartsWith("+*");
         }
 
-        List<phonemesPerNote> ProcessWord(Note[] notes){
+        /// <summary>
+        /// distribute phonemes to each note inside the group
+        /// </summary>
+        List<phonemesPerNote> ProcessWord(Note[] notes, string[] symbols){
             var wordPhonemes = new List<phonemesPerNote>{
                 new phonemesPerNote(-1, notes[0].tone)
             };
-            var dsPhonemes = GetDsPhonemes(notes[0]);
+            var dsPhonemes = symbols
+                .Select((symbol, index) => new dsPhoneme(symbol, GetSpeakerAtIndex(notes[0], index)))
+                .ToArray(); 
             var isVowel = dsPhonemes.Select(s => g2p.IsVowel(s.Symbol)).ToArray();
             var isGlide = dsPhonemes.Select(s => g2p.IsGlide(s.Symbol)).ToArray();
             var nonExtensionNotes = notes.Where(n=>!IsSyllableVowelExtensionNote(n)).ToArray();
@@ -217,8 +216,17 @@ protected override void ProcessPart(Note[][] phrase) {
                 new phonemesPerNote(-1,phrase[0][0].tone, new List<dsPhoneme>{new dsPhoneme("SP", GetSpeakerAtIndex(phrase[0][0], 0))})
             };
             var notePhIndex = new List<int> { 1 };
-            foreach (var word in phrase) {
-                var wordPhonemes = ProcessWord(word);
+            var wordFound = new bool[phrase.Length];
+            foreach (int wordIndex in Enumerable.Range(0, phrase.Length)) {
+                Note[] word = phrase[wordIndex];
+                var symbols = GetSymbols(word[0]);
+                if (symbols == null || symbols.Length == 0) {
+                    symbols = new string[] { defaultPause };
+                    wordFound[wordIndex] = false;
+                } else {
+                    wordFound[wordIndex] = true;
+                }
+                var wordPhonemes = ProcessWord(word, symbols);
                 phrasePhonemes[^1].Phonemes.AddRange(wordPhonemes[0].Phonemes);
                 phrasePhonemes.AddRange(wordPhonemes.Skip(1));
                 notePhIndex.Add(notePhIndex[^1]+wordPhonemes.SelectMany(n=>n.Phonemes).Count());
@@ -310,20 +318,24 @@ protected override void ProcessPart(Note[][] phrase) {
 
             //Convert the position sequence to tick and fill into the result list
             int index = 1;
-            foreach (int groupIndex in Enumerable.Range(0, phrase.Length)) {
-                Note[] group = phrase[groupIndex];
+            foreach (int wordIndex in Enumerable.Range(0, phrase.Length)) {
+                Note[] word = phrase[wordIndex];
                 var noteResult = new List<Tuple<string, int>>();
-                if (group[0].lyric.StartsWith("+")) {
+                if (!wordFound[wordIndex]){
+                    //partResult[word[0].position] = noteResult;
+                    continue;
+                }
+                if (word[0].lyric.StartsWith("+")) {
                     continue;
                 }
-                double notePos = timeAxis.TickPosToMsPos(group[0].position);//start position of the note, ms
-                for (int phIndex = notePhIndex[groupIndex]; phIndex < notePhIndex[groupIndex + 1]; ++phIndex) {
+                double notePos = timeAxis.TickPosToMsPos(word[0].position);//start position of the note, ms
+                for (int phIndex = notePhIndex[wordIndex]; phIndex < notePhIndex[wordIndex + 1]; ++phIndex) {
                     if (!String.IsNullOrEmpty(phs[phIndex].Symbol)) {
                         noteResult.Add(Tuple.Create(phs[phIndex].Symbol, timeAxis.TicksBetweenMsPos(
                            notePos, positions[phIndex - 1])));
                     }
                 }
-                partResult[group[0].position] = noteResult;
+                partResult[word[0].position] = noteResult;
             }
         }
     }

diff --git a/OpenUtau.Core/DiffSinger/DiffSingerConfig.cs b/OpenUtau.Core/DiffSinger/DiffSingerConfig.cs
@@ -23,7 +23,9 @@ public class DsConfig {
         public bool useKeyShiftEmbed = false;
         public bool useSpeedEmbed = false;
         public bool useEnergyEmbed = false;
-        public bool useBreathinessEmbed= false;
+        public bool useBreathinessEmbed = false;
+        public bool useVoicingEmbed = false;
+        public bool useTensionEmbed = false;
         public AugmentationArgs augmentationArgs;
         public bool useShallowDiffusion = false;
         public int maxDepth = -1;
@@ -34,6 +36,10 @@ public class DsConfig {
         public int hop_size = 512;
         public int sample_rate = 44100;
         public bool predict_dur = true;
+        public bool predict_energy = true;
+        public bool predict_breathiness = true;
+        public bool predict_voicing = false;
+        public bool predict_tension = false;
         public bool use_expr = false;
         public bool use_note_rest = false;
         public float frameMs(){

diff --git a/OpenUtau.Core/DiffSinger/DiffSingerPitch.cs b/OpenUtau.Core/DiffSinger/DiffSingerPitch.cs
@@ -116,7 +116,7 @@ public RenderPitchResult Process(RenderPhrase phrase){
                     new DenseTensor<Int64>(word_dur, new int[] { word_dur.Length }, false)
                     .Reshape(new int[] { 1, word_dur.Length })));
             }else{
-                //if predict_dur is true, use phoneme encode mode
+                //if predict_dur is false, use phoneme encode mode
                 linguisticInputs.Add(NamedOnnxValue.CreateFromTensor("ph_dur",
                     new DenseTensor<Int64>(ph_dur.Select(x=>(Int64)x).ToArray(), new int[] { ph_dur.Length }, false)
                     .Reshape(new int[] { 1, ph_dur.Length })));