com.rest.elevenlabs 3.2.0 (#52)

- updated com.utilities.rest -> 2.4.0 - fixed text to speech request encoding if test is encoded to something other than utf8
RageAgainstThePixel · Dec 14, 2023 · 19515c4 · 19515c4
1 parent 75f4390
commit 19515c4
Show file tree

Hide file tree

Showing 6 changed files with 16 additions and 10 deletions.
diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/History/HistoryEndpoint.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/History/HistoryEndpoint.cs
@@ -113,7 +113,7 @@ public async Task<VoiceClip> DownloadHistoryAudioAsync(HistoryItem historyItem,
             }
 
             await Awaiters.UnityMainThread;
-            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", AudioType.UNKNOWN, cancellationToken: cancellationToken);
+            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", AudioType.UNKNOWN, debug: EnableDebug, cancellationToken: cancellationToken);
             return new VoiceClip(historyItem.Id, historyItem.Text, voice, audioClip, cachedPath);
         }
 

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/TextToSpeechEndpoint.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/TextToSpeechEndpoint.cs
@@ -66,7 +66,6 @@ public async Task<VoiceClip> TextToSpeechAsync(string text, Voice voice, VoiceSe
         {
             ValidateInputs(text, voice);
 
-            var downloadDirectory = await GetCacheDirectoryAsync(voice);
             var defaultVoiceSettings = voiceSettings ?? voice.Settings ?? await client.VoicesEndpoint.GetDefaultVoiceSettingsAsync(cancellationToken);
             var request = new TextToSpeechRequest(text, model, defaultVoiceSettings);
             var payload = JsonConvert.SerializeObject(request, ElevenLabsClient.JsonSerializationOptions);
@@ -95,6 +94,7 @@ public async Task<VoiceClip> TextToSpeechAsync(string text, Voice voice, VoiceSe
                 AudioType.OGGVORBIS => "ogg",
                 _ => throw new ArgumentOutOfRangeException($"Unsupported {nameof(AudioType)}: {audioType}")
             };
+            var downloadDirectory = await GetCacheDirectoryAsync(voice);
             var cachedPath = $"{downloadDirectory}/{clipId}.{extension}";
 
             if (!File.Exists(cachedPath))
@@ -115,15 +115,15 @@ public async Task<VoiceClip> TextToSpeechAsync(string text, Voice voice, VoiceSe
                             _ => throw new ArgumentOutOfRangeException(nameof(outputFormat), outputFormat, null)
                         };
                         var oggBytes = await OggEncoder.ConvertToBytesAsync(pcmData, frequency, 1, cancellationToken: cancellationToken).ConfigureAwait(false);
-                        await File.WriteAllBytesAsync(cachedPath, oggBytes, cancellationToken: cancellationToken).ConfigureAwait(false);
+                        await File.WriteAllBytesAsync(cachedPath, oggBytes, cancellationToken).ConfigureAwait(false);
                         break;
                     default:
                         throw new ArgumentOutOfRangeException($"Unsupported {nameof(AudioType)}: {audioType}");
                 }
             }
 
             await Awaiters.UnityMainThread;
-            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", audioType, cancellationToken: cancellationToken);
+            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", audioType, debug: EnableDebug, cancellationToken: cancellationToken);
             return new VoiceClip(clipId, text, voice, audioClip, cachedPath);
         }
 
@@ -182,7 +182,6 @@ public async Task<VoiceClip> StreamTextToSpeechAsync(string text, Voice voice, A
                 OutputFormat.PCM_44100 => 44100,
                 _ => throw new ArgumentOutOfRangeException(nameof(outputFormat), outputFormat, null)
             };
-            var downloadDirectory = await GetCacheDirectoryAsync(voice);
             var defaultVoiceSettings = voiceSettings ?? voice.Settings ?? await client.VoicesEndpoint.GetDefaultVoiceSettingsAsync(cancellationToken);
             var request = new TextToSpeechRequest(text, model, defaultVoiceSettings);
             var payload = JsonConvert.SerializeObject(request, ElevenLabsClient.JsonSerializationOptions);
@@ -208,10 +207,11 @@ public async Task<VoiceClip> StreamTextToSpeechAsync(string text, Voice voice, A
             var pcmData = PCMEncoder.Decode(response.Data, PCMFormatSize.SixteenBit);
             var fullClip = AudioClip.Create(clipId, pcmData.Length, 1, frequency, false);
             var oggBytes = await OggEncoder.ConvertToBytesAsync(pcmData, frequency, 1, cancellationToken: cancellationToken).ConfigureAwait(false);
+            var downloadDirectory = await GetCacheDirectoryAsync(voice);
             var cachedPath = $"{downloadDirectory}/{clipId}.ogg";
             await File.WriteAllBytesAsync(cachedPath, oggBytes, cancellationToken: cancellationToken).ConfigureAwait(false);
             await Awaiters.UnityMainThread;
-            await Rest.DownloadAudioClipAsync($"file://{cachedPath}", AudioType.OGGVORBIS, cancellationToken: cancellationToken);
+            await Rest.DownloadAudioClipAsync($"file://{cachedPath}", AudioType.OGGVORBIS, debug: EnableDebug, cancellationToken: cancellationToken);
             return new VoiceClip(clipId, text, voice, fullClip, cachedPath);
 
             void StreamCallback(Response partialResponse)

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/TextToSpeechRequest.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/TextToSpeech/TextToSpeechRequest.cs
@@ -4,6 +4,7 @@
 using ElevenLabs.Voices;
 using Newtonsoft.Json;
 using System;
+using System.Text;
 using UnityEngine.Scripting;
 
 namespace ElevenLabs.TextToSpeech
@@ -22,6 +23,11 @@ internal sealed class TextToSpeechRequest
                 throw new ArgumentNullException(nameof(text));
             }
 
+            if (!Encoding.GetEncoding(text).Equals(Encoding.UTF8))
+            {
+                text = Encoding.UTF8.GetString(Encoding.Default.GetBytes(text));
+            }
+
             Text = text;
             Model = model ?? Models.Model.MonoLingualV1;
             VoiceSettings = voiceSettings ?? throw new ArgumentNullException(nameof(voiceSettings));

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/VoiceGeneration/VoiceGenerationEndpoint.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/VoiceGeneration/VoiceGenerationEndpoint.cs
@@ -55,7 +55,7 @@ public async Task<GeneratedVoiceOptions> GetVoiceGenerationOptionsAsync(Cancella
             }
 
             await File.WriteAllBytesAsync(cachedPath, response.Data, cancellationToken).ConfigureAwait(true);
-            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", AudioType.MPEG, cancellationToken: cancellationToken);
+            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", AudioType.MPEG, debug: EnableDebug, cancellationToken: cancellationToken);
             return new Tuple<string, AudioClip>(generatedVoiceId, audioClip);
         }
 

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/Voices/VoicesEndpoint.cs b/ElevenLabs/Packages/com.rest.elevenlabs/Runtime/Voices/VoicesEndpoint.cs
@@ -339,7 +339,7 @@ public async Task<VoiceClip> DownloadVoiceSampleAudioAsync(Voice voice, Sample s
                 }
             }
 
-            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", audioType, cancellationToken: cancellationToken);
+            var audioClip = await Rest.DownloadAudioClipAsync($"file://{cachedPath}", audioType, debug: EnableDebug, cancellationToken: cancellationToken);
             return new VoiceClip(sample.Id, string.Empty, voice, audioClip, cachedPath);
         }
 

diff --git a/ElevenLabs/Packages/com.rest.elevenlabs/package.json b/ElevenLabs/Packages/com.rest.elevenlabs/package.json
@@ -3,7 +3,7 @@
   "displayName": "ElevenLabs",
   "description": "A non-official Eleven Labs voice synthesis RESTful client.",
   "keywords": [],
-  "version": "3.1.2",
+  "version": "3.2.0",
   "unity": "2021.3",
   "documentationUrl": "https://github.com/RageAgainstThePixel/com.rest.elevenlabs#documentation",
   "changelogUrl": "https://github.com/RageAgainstThePixel/com.rest.elevenlabs/releases",
@@ -17,7 +17,7 @@
     "url": "https://github.com/StephenHodgson"
   },
   "dependencies": {
-    "com.utilities.rest": "2.3.1",
+    "com.utilities.rest": "2.4.0",
     "com.utilities.encoder.ogg": "3.0.12"
   },
   "samples": [