[Fix] TTS (#137)

* TTS fix * TTS fix * TTS fix * RedFoxIV revie
WWhiteDreamProject · Dec 5, 2024 · c5bd6b7 · c5bd6b7
1 parent 8267e36
commit c5bd6b7
Show file tree

Hide file tree

Showing 4 changed files with 38 additions and 49 deletions.
diff --git a/Content.Server/_White/TTS/TTSManager.cs b/Content.Server/_White/TTS/TTSManager.cs
@@ -50,7 +50,7 @@ public void Initialize()
     /// <param name="speaker">Identifier of speaker</param>
     /// <param name="text">SSML formatted text</param>
     /// <returns>OGG audio bytes</returns>
-    public async Task<byte[]?> ConvertTextToSpeech(string speaker, string text, string pitch, string rate, string? effect = null)
+    public async Task<byte[]?> ConvertTextToSpeech(string speaker, string text)
     {
         var url = _cfg.GetCVar(WhiteCVars.TTSApiUrl);
         if (string.IsNullOrWhiteSpace(url))
@@ -80,10 +80,7 @@ public void Initialize()
         {
             ApiToken = token,
             Text = text,
-            Speaker = speaker,
-            Pitch = pitch,
-            Rate = rate,
-            Effect = effect
+            Speaker = speaker
         };
 
         var reqTime = DateTime.UtcNow;
@@ -143,19 +140,28 @@ private record GenerateVoiceRequest
         public string ApiToken { get; set; } = "";
 
         [JsonPropertyName("text")]
-        public string Text { get; set; } = default!;
+        public string Text { get; set; } = "";
 
         [JsonPropertyName("speaker")]
-        public string Speaker { get; set; } = default!;
+        public string Speaker { get; set; } = "";
 
-        [JsonPropertyName("pitch")]
-        public string Pitch { get; set; } = default!;
+        [JsonPropertyName("ssml")]
+        public bool SSML { get; private set; } = true;
 
-        [JsonPropertyName("rate")]
-        public string Rate { get; set; } = default!;
+        [JsonPropertyName("word_ts")]
+        public bool WordTS { get; private set; } = false;
 
-        [JsonPropertyName("effect")]
-        public string? Effect { get; set; }
+        [JsonPropertyName("put_accent")]
+        public bool PutAccent { get; private set; } = true;
+
+        [JsonPropertyName("put_yo")]
+        public bool PutYo { get; private set; } = false;
+
+        [JsonPropertyName("sample_rate")]
+        public int SampleRate { get; private set; } = 24000;
+
+        [JsonPropertyName("format")]
+        public string Format { get; private set; } = "ogg";
     }
 
     private struct GenerateVoiceResponse

diff --git a/Content.Server/_White/TTS/TTSSystem.cs b/Content.Server/_White/TTS/TTSSystem.cs
@@ -113,7 +113,7 @@ private async void OnAnnounceRequest(TtsAnnouncementEvent ev)
         if (!_prototypeManager.TryIndex(ev.VoiceId, out var ttsPrototype))
             return;
         var message = FormattedMessage.RemoveMarkup(ev.Message);
-        var soundData = await GenerateTTS(null, message, ttsPrototype.Speaker, speechRate: "slow", effect: "announce");
+        var soundData = await GenerateTTS(null, message, ttsPrototype.Speaker, speechRate: "slow");
         if (soundData == null)
             return;
         Filter filter;
@@ -184,35 +184,14 @@ private async void OnRequestTTS(MsgRequestTTS ev)
             RaiseNetworkEvent(new PlayTTSEvent(ev.Uid, soundData, false), Filter.SinglePlayer(session), false);
     }
 
-    private async Task<byte[]?> GenerateTTS(EntityUid? uid, string text, string speaker, string? speechPitch = null,
-        string? speechRate = null, string? effect = null)
+    private async Task<byte[]?> GenerateTTS(EntityUid? uid, string text, string speaker, string? speechRate = null, string? speechPitch = null)
     {
         var textSanitized = Sanitize(text);
         if (textSanitized == "")
             return null;
 
-        string pitch;
-        string rate;
-        if (speechPitch == null || speechRate == null)
-        {
-            if (uid == null || !_ttsPitchRateSystem.TryGetPitchRate(uid.Value, out var pitchRate))
-            {
-                pitch = "medium";
-                rate = "medium";
-            }
-            else
-            {
-                pitch = pitchRate.Pitch;
-                rate = pitchRate.Rate;
-            }
-        }
-        else
-        {
-            pitch = speechPitch;
-            rate = speechRate;
-        }
-
-        return await _ttsManager.ConvertTextToSpeech(speaker, textSanitized, pitch, rate, effect);
+        textSanitized = _ttsPitchRateSystem.GetFormattedSpeechText(uid, textSanitized, speechRate, speechPitch);
+        return await _ttsManager.ConvertTextToSpeech(speaker, textSanitized);
     }
 }
 

diff --git a/Content.Shared/_White/TTS/TTSPitchRateSystem.cs b/Content.Shared/_White/TTS/TTSPitchRateSystem.cs
@@ -1,4 +1,3 @@
-using System.Diagnostics.CodeAnalysis;
 using Content.Shared.Humanoid;
 using Content.Shared.Humanoid.Prototypes;
 using Robust.Shared.Prototypes;
@@ -18,21 +17,25 @@ public sealed class TTSPitchRateSystem : EntitySystem
         ["Reptilian"] = new TTSPitchRate("low", "slow"),
     };
 
-    public bool TryGetPitchRate(EntityUid uid, [NotNullWhen(true)] out TTSPitchRate? pitch)
+    public string GetFormattedSpeechText(EntityUid? uid, string text, string? speechRate = null, string? speechPitch = null)
     {
-        if (!TryComp<HumanoidAppearanceComponent>(uid, out var humanoid))
+        var ssml = text;
+        if (TryComp<HumanoidAppearanceComponent>(uid, out var humanoid))
         {
-            pitch = new TTSPitchRate();
-            return false;
+            var species = SpeciesPitches.GetValueOrDefault(humanoid.Species);
+            if (species != null)
+            {
+                speechRate ??= species.Rate;
+                speechPitch ??= species.Pitch;
+            }
         }
 
-        pitch = GetPitchRate(humanoid.Species);
-        return pitch != null;
-    }
+        if (speechRate != null)
+            ssml = $"<prosody rate=\"{speechRate}\">{ssml}</prosody>";
+        if (speechPitch != null)
+            ssml = $"<prosody pitch=\"{speechPitch}\">{ssml}</prosody>";
 
-    public TTSPitchRate? GetPitchRate(ProtoId<SpeciesPrototype> protoId)
-    {
-        return SpeciesPitches.GetValueOrDefault(protoId);
+        return $"<speak>{ssml}</speak>";
     }
 }
 

diff --git a/Resources/Prototypes/Entities/Mobs/Species/base.yml b/Resources/Prototypes/Entities/Mobs/Species/base.yml
@@ -230,6 +230,7 @@
     - DoorBumpOpener
   - type: Targeting
   - type: SurgeryTarget
+  - type: TTS
 
 - type: entity
   save: false