added unit tests

RageAgainstThePixel · Nov 24, 2024 · eda0bdb · eda0bdb
1 parent a354f84
commit eda0bdb
Show file tree

Hide file tree

Showing 3 changed files with 56 additions and 1 deletion.
diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/Alignment.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/Alignment.cs
@@ -35,6 +35,7 @@ internal Alignment(
         [Preserve]
         public static implicit operator TimestampedTranscriptCharacter[](Alignment alignment)
         {
+            if (alignment == null) { return null; }
             var characters = alignment.Characters;
             var startTimes = alignment.StartTimes;
             var endTimes = alignment.EndTimes;

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/TextToSpeechEndpoint.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/TextToSpeechEndpoint.cs
@@ -431,7 +431,7 @@ void ProcessPartialResponse(string line)
                         return;
                     }
 
-                    var timestampedTranscriptCharacters = (TimestampedTranscriptCharacter[])partialResponse.Alignment;
+                    var timestampedTranscriptCharacters = (TimestampedTranscriptCharacter[])partialResponse.Alignment ?? Array.Empty<TimestampedTranscriptCharacter>();
                     partialClipCallback.Invoke(Tuple.Create(audioClip, timestampedTranscriptCharacters));
                     accumulatedPCMData.AddRange(partialClipData);
                     accumulatedTranscriptData.AddRange(timestampedTranscriptCharacters);

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Tests/Test_Fixture_04_TextToSpeechEndpoint.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Tests/Test_Fixture_04_TextToSpeechEndpoint.cs
@@ -1,6 +1,7 @@
 // Licensed under the MIT License. See LICENSE in the project root for license information.
 
 using NUnit.Framework;
+using System;
 using System.Collections.Generic;
 using System.Linq;
 using System.Threading.Tasks;
@@ -40,5 +41,58 @@ public async Task Test_02_StreamTextToSpeech()
             Assert.IsNotNull(voiceClip.AudioClip);
             Debug.Log(voiceClip.Id);
         }
+
+        [Test]
+        public async Task Test_03_TextToSpeech_Transcription()
+        {
+            Assert.NotNull(ElevenLabsClient.TextToSpeechEndpoint);
+            var voice = Voices.Voice.Adam;
+            Assert.NotNull(voice);
+            var defaultVoiceSettings = await ElevenLabsClient.VoicesEndpoint.GetDefaultVoiceSettingsAsync();
+            var voiceClip = await ElevenLabsClient.TextToSpeechEndpoint.TextToSpeechWithTimestampsAsync("The quick brown fox jumps over the lazy dog.", voice, defaultVoiceSettings);
+            Assert.NotNull(voiceClip);
+            Assert.NotNull(voiceClip.AudioClip);
+            Debug.Log(voiceClip.Id);
+            Assert.NotNull(voiceClip.TimestampedTranscriptCharacters);
+            Assert.IsNotEmpty(voiceClip.TimestampedTranscriptCharacters);
+            Debug.Log("| Character | Start Time | End Time |");
+            Debug.Log("| --------- | ---------- | -------- |");
+            foreach (var character in voiceClip.TimestampedTranscriptCharacters)
+            {
+                Debug.Log($"| {character.Character} | {character.StartTime} | {character.EndTime} |");
+            }
+        }
+
+        [Test]
+        public async Task Test_04_StreamTextToSpeech_Transcription()
+        {
+            Assert.NotNull(ElevenLabsClient.TextToSpeechEndpoint);
+            var voice = Voices.Voice.Adam;
+            Assert.NotNull(voice);
+            var defaultVoiceSettings = await ElevenLabsClient.VoicesEndpoint.GetDefaultVoiceSettingsAsync();
+            var partialClips = new Queue<AudioClip>();
+            var characters = new Queue<TimestampedTranscriptCharacter>();
+            Debug.Log("| Character | Start Time | End Time |");
+            Debug.Log("| --------- | ---------- | -------- |");
+            var voiceClip = await ElevenLabsClient.TextToSpeechEndpoint.StreamTextToSpeechWithTimestampsAsync("The quick brown fox jumps over the lazy dog.", voice,
+                 callback =>
+                 {
+                     var (partialClip, chars) = callback;
+                     partialClips.Enqueue(partialClip);
+                     foreach (var character in chars)
+                     {
+                         characters.Enqueue(character);
+                         Debug.Log($"| {character.Character} | {character.StartTime} | {character.EndTime} |");
+                     }
+                 },
+                 defaultVoiceSettings);
+            Assert.NotNull(partialClips);
+            Assert.NotNull(partialClips);
+            Assert.IsNotEmpty(partialClips);
+            Assert.NotNull(voiceClip);
+            Assert.IsNotNull(voiceClip.AudioClip);
+            Debug.Log(voiceClip.Id);
+            Assert.AreEqual(characters.ToArray(), voiceClip.TimestampedTranscriptCharacters);
+        }
     }
 }