From fb4971d8e911d2dc166f77cbd53d5ada03f6eae1 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 04:48:33 -0800
Subject: [PATCH 01/12] ...

---
 lib/classes/tts_engines/bark.py     |  7 +++++--
 lib/classes/tts_engines/fairseq.py  |  7 +++++--
 lib/classes/tts_engines/tacotron.py |  7 +++++--
 lib/classes/tts_engines/vits.py     |  7 +++++--
 lib/classes/tts_engines/xtts.py     | 10 +++++-----
 lib/classes/tts_engines/yourtts.py  |  7 +++++--
 6 files changed, 30 insertions(+), 15 deletions(-)

diff --git a/lib/classes/tts_engines/bark.py b/lib/classes/tts_engines/bark.py
index 92c5f294..358e19ac 100644
--- a/lib/classes/tts_engines/bark.py
+++ b/lib/classes/tts_engines/bark.py
@@ -197,8 +197,11 @@ class Bark(TTSUtils, TTSRegistry, name='bark'):
                     #if is_audio_data_valid(audio_sentence):
                     #    audio_sentence = audio_sentence.tolist()
                     if is_audio_data_valid(audio_sentence):
-                        sourceTensor = self._tensor_type(audio_sentence)
-                        audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
+                        if not isinstance(audio_sentence, torch.Tensor):
+                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                            print(error)
+                            return False
+                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/fairseq.py b/lib/classes/tts_engines/fairseq.py
index 91637f42..e5cb1562 100644
--- a/lib/classes/tts_engines/fairseq.py
+++ b/lib/classes/tts_engines/fairseq.py
@@ -166,8 +166,11 @@ class Fairseq(TTSUtils, TTSRegistry, name='fairseq'):
                                 **speaker_argument
                             )
                     if is_audio_data_valid(audio_sentence):
-                        sourceTensor = self._tensor_type(audio_sentence)
-                        audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
+                        if not isinstance(audio_sentence, torch.Tensor):
+                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                            print(error)
+                            return False
+                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/tacotron.py b/lib/classes/tts_engines/tacotron.py
index 3238de37..5156f88f 100644
--- a/lib/classes/tts_engines/tacotron.py
+++ b/lib/classes/tts_engines/tacotron.py
@@ -194,8 +194,11 @@ class Tacotron2(TTSUtils, TTSRegistry, name='tacotron'):
                                 **speaker_argument
                             )
                     if is_audio_data_valid(audio_sentence):
-                        sourceTensor = self._tensor_type(audio_sentence)
-                        audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
+                        if not isinstance(audio_sentence, torch.Tensor):
+                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                            print(error)
+                            return False
+                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/vits.py b/lib/classes/tts_engines/vits.py
index 7d6654b7..ec7b40b2 100644
--- a/lib/classes/tts_engines/vits.py
+++ b/lib/classes/tts_engines/vits.py
@@ -179,8 +179,11 @@ class Vits(TTSUtils, TTSRegistry, name='vits'):
                                 **speaker_argument
                             )
                     if is_audio_data_valid(audio_sentence):
-                        sourceTensor = self._tensor_type(audio_sentence)
-                        audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
+                        if not isinstance(audio_sentence, torch.Tensor):
+                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                            print(error)
+                            return False
+                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/xtts.py b/lib/classes/tts_engines/xtts.py
index 0c9beaa7..9f0fe864 100644
--- a/lib/classes/tts_engines/xtts.py
+++ b/lib/classes/tts_engines/xtts.py
@@ -140,11 +140,11 @@ class XTTSv2(TTSUtils, TTSRegistry, name='xtts'):
                             **fine_tuned_params
                         )
                     audio_sentence = result.get('wav')
-                    if is_audio_data_valid(audio_sentence):
-                        audio_sentence = audio_sentence.tolist()
-                    if is_audio_data_valid(audio_sentence):
-                        sourceTensor = self._tensor_type(audio_sentence)
-                        audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
+                    if not isinstance(audio_sentence, torch.Tensor):
+                        error = 'XTTSv2 inference did not return a torch.Tensor'
+                        print(error)
+                        return False
+                    audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/yourtts.py b/lib/classes/tts_engines/yourtts.py
index e185e5d4..713d9352 100644
--- a/lib/classes/tts_engines/yourtts.py
+++ b/lib/classes/tts_engines/yourtts.py
@@ -109,8 +109,11 @@ class YourTTS(TTSUtils, TTSRegistry, name='yourtts'):
                             **speaker_argument
                         )
                     if is_audio_data_valid(audio_sentence):
-                        sourceTensor = self._tensor_type(audio_sentence)
-                        audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
+                        if not isinstance(audio_sentence, torch.Tensor):
+                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                            print(error)
+                            return False
+                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:

From cac7cab38e0ca0269b93e35f5a85768b9dfa092a Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 04:54:40 -0800
Subject: [PATCH 02/12] ...

---
 lib/classes/tts_engines/xtts.py | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/lib/classes/tts_engines/xtts.py b/lib/classes/tts_engines/xtts.py
index 9f0fe864..ea20f44c 100644
--- a/lib/classes/tts_engines/xtts.py
+++ b/lib/classes/tts_engines/xtts.py
@@ -140,13 +140,14 @@ class XTTSv2(TTSUtils, TTSRegistry, name='xtts'):
                             **fine_tuned_params
                         )
                     audio_sentence = result.get('wav')
-                    if not isinstance(audio_sentence, torch.Tensor):
-                        error = 'XTTSv2 inference did not return a torch.Tensor'
-                        print(error)
-                        return False
-                    audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
-                        if sentence[-1].isalnum() or sentence[-1] == '—':
-                            audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
+                    if is_audio_data_valid(audio_sentence):
+                        if not isinstance(audio_sentence, torch.Tensor):
+                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                            print(error)
+                            return False
+                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                            if sentence[-1].isalnum() or sentence[-1] == '—':
+                                audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
                             self.audio_segments.append(audio_tensor)
                             if not re.search(r'\w$', sentence, flags=re.UNICODE) and sentence[-1] != '—':

From 263e018792ff667fa0ba58a3261edec6721b375e Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 04:56:31 -0800
Subject: [PATCH 03/12] ...

---
 lib/classes/tts_engines/xtts.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/lib/classes/tts_engines/xtts.py b/lib/classes/tts_engines/xtts.py
index ea20f44c..5de98832 100644
--- a/lib/classes/tts_engines/xtts.py
+++ b/lib/classes/tts_engines/xtts.py
@@ -178,6 +178,10 @@ class XTTSv2(TTSUtils, TTSRegistry, name='xtts'):
                                 error = f"Cannot create {final_sentence_file}"
                                 print(error)
                                 return False
+                        else:
+                            error = f"audio_tensor not valid"
+                            print(error)
+                            return False
                     else:
                         error = f"audio_sentence not valid"
                         print(error)

From a0191f9b6cbb73a784f4525d762f3bb73c0a2108 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 04:58:58 -0800
Subject: [PATCH 04/12] ...

---
 lib/classes/tts_engines/bark.py     | 4 ++++
 lib/classes/tts_engines/fairseq.py  | 4 ++++
 lib/classes/tts_engines/tacotron.py | 4 ++++
 lib/classes/tts_engines/vits.py     | 4 ++++
 lib/classes/tts_engines/xtts.py     | 4 ++--
 lib/classes/tts_engines/yourtts.py  | 4 ++++
 6 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/lib/classes/tts_engines/bark.py b/lib/classes/tts_engines/bark.py
index 358e19ac..6a85b27f 100644
--- a/lib/classes/tts_engines/bark.py
+++ b/lib/classes/tts_engines/bark.py
@@ -234,6 +234,10 @@ class Bark(TTSUtils, TTSRegistry, name='bark'):
                                 error = f"Cannot create {final_sentence_file}"
                                 print(error)
                                 return False
+                        else:
+                            error = f"audio_tensor not valid"
+                            print(error)
+                            return False
                     else:
                         error = f"audio_sentence not valid"
                         print(error)
diff --git a/lib/classes/tts_engines/fairseq.py b/lib/classes/tts_engines/fairseq.py
index e5cb1562..c9759f7f 100644
--- a/lib/classes/tts_engines/fairseq.py
+++ b/lib/classes/tts_engines/fairseq.py
@@ -203,6 +203,10 @@ class Fairseq(TTSUtils, TTSRegistry, name='fairseq'):
                                 error = f"Cannot create {final_sentence_file}"
                                 print(error)
                                 return False
+                        else:
+                            error = f"audio_tensor not valid"
+                            print(error)
+                            return False
                     else:
                         error = f"audio_sentence not valid"
                         print(error)
diff --git a/lib/classes/tts_engines/tacotron.py b/lib/classes/tts_engines/tacotron.py
index 5156f88f..3285a9b3 100644
--- a/lib/classes/tts_engines/tacotron.py
+++ b/lib/classes/tts_engines/tacotron.py
@@ -231,6 +231,10 @@ class Tacotron2(TTSUtils, TTSRegistry, name='tacotron'):
                                 error = f"Cannot create {final_sentence_file}"
                                 print(error)
                                 return False
+                        else:
+                            error = f"audio_tensor not valid"
+                            print(error)
+                            return False
                     else:
                         error = f"audio_sentence not valid"
                         print(error)
diff --git a/lib/classes/tts_engines/vits.py b/lib/classes/tts_engines/vits.py
index ec7b40b2..09cd7f79 100644
--- a/lib/classes/tts_engines/vits.py
+++ b/lib/classes/tts_engines/vits.py
@@ -216,6 +216,10 @@ class Vits(TTSUtils, TTSRegistry, name='vits'):
                                 error = f"Cannot create {final_sentence_file}"
                                 print(error)
                                 return False
+                        else:
+                            error = f"audio_tensor not valid"
+                            print(error)
+                            return False
                     else:
                         error = f"audio_sentence not valid"
                         print(error)
diff --git a/lib/classes/tts_engines/xtts.py b/lib/classes/tts_engines/xtts.py
index 5de98832..9eec3348 100644
--- a/lib/classes/tts_engines/xtts.py
+++ b/lib/classes/tts_engines/xtts.py
@@ -146,8 +146,8 @@ class XTTSv2(TTSUtils, TTSRegistry, name='xtts'):
                             print(error)
                             return False
                         audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
-                            if sentence[-1].isalnum() or sentence[-1] == '—':
-                                audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
+                        if sentence[-1].isalnum() or sentence[-1] == '—':
+                            audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
                             self.audio_segments.append(audio_tensor)
                             if not re.search(r'\w$', sentence, flags=re.UNICODE) and sentence[-1] != '—':
diff --git a/lib/classes/tts_engines/yourtts.py b/lib/classes/tts_engines/yourtts.py
index 713d9352..b049f07f 100644
--- a/lib/classes/tts_engines/yourtts.py
+++ b/lib/classes/tts_engines/yourtts.py
@@ -146,6 +146,10 @@ class YourTTS(TTSUtils, TTSRegistry, name='yourtts'):
                                 error = f"Cannot create {final_sentence_file}"
                                 print(error)
                                 return False
+                        else:
+                            error = f"audio_tensor not valid"
+                            print(error)
+                            return False
                     else:
                         error = f"audio_sentence not valid"
                         print(error)

From 002e4f415b9d6e0487b88455e1890b5f626a75c8 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 10:39:47 -0800
Subject: [PATCH 05/12] ...

---
 lib/classes/tts_engines/xtts.py | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/lib/classes/tts_engines/xtts.py b/lib/classes/tts_engines/xtts.py
index 9eec3348..fc8a4b21 100644
--- a/lib/classes/tts_engines/xtts.py
+++ b/lib/classes/tts_engines/xtts.py
@@ -141,11 +141,16 @@ class XTTSv2(TTSUtils, TTSRegistry, name='xtts'):
                         )
                     audio_sentence = result.get('wav')
                     if is_audio_data_valid(audio_sentence):
-                        if not isinstance(audio_sentence, torch.Tensor):
-                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported XTTSv2 wav type: {type(audio_sentence)}"
                             print(error)
                             return False
-                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:

From b00bf24782200cbdc6a3b93674b1392b292b2471 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 10:41:23 -0800
Subject: [PATCH 06/12] ...

---
 lib/classes/tts_engines/common/utils.py | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/lib/classes/tts_engines/common/utils.py b/lib/classes/tts_engines/common/utils.py
index a297ddb6..4fae9aaf 100644
--- a/lib/classes/tts_engines/common/utils.py
+++ b/lib/classes/tts_engines/common/utils.py
@@ -207,7 +207,16 @@ class TTSUtils:
                                 speaker_embedding=speaker_embedding,
                                 **fine_tuned_params,
                             )
-                        audio_sentence = result.get('wav') if isinstance(result, dict) else None
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported XTTSv2 wav type: {type(audio_sentence)}"
+                            print(error)
+                            return False
                         if audio_sentence is not None:
                             audio_sentence = audio_sentence.tolist()
                             sourceTensor = self._tensor_type(audio_sentence)

From 5ad336b279c1c9d17639900a896cb146e1f3a5b3 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 12:11:17 -0800
Subject: [PATCH 07/12] ...

---
 lib/classes/tts_engines/bark.py     | 10 ++++++++--
 lib/classes/tts_engines/fairseq.py  | 11 ++++++++---
 lib/classes/tts_engines/tacotron.py | 11 ++++++++---
 lib/classes/tts_engines/vits.py     | 11 ++++++++---
 lib/classes/tts_engines/yourtts.py  | 11 ++++++++---
 5 files changed, 40 insertions(+), 14 deletions(-)

diff --git a/lib/classes/tts_engines/bark.py b/lib/classes/tts_engines/bark.py
index 6a85b27f..a149c88f 100644
--- a/lib/classes/tts_engines/bark.py
+++ b/lib/classes/tts_engines/bark.py
@@ -198,10 +198,16 @@ class Bark(TTSUtils, TTSRegistry, name='bark'):
                     #    audio_sentence = audio_sentence.tolist()
                     if is_audio_data_valid(audio_sentence):
                         if not isinstance(audio_sentence, torch.Tensor):
-                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported Bark wav type: {type(audio_sentence)}"
                             print(error)
                             return False
-                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/fairseq.py b/lib/classes/tts_engines/fairseq.py
index c9759f7f..9ae232d9 100644
--- a/lib/classes/tts_engines/fairseq.py
+++ b/lib/classes/tts_engines/fairseq.py
@@ -166,11 +166,16 @@ class Fairseq(TTSUtils, TTSRegistry, name='fairseq'):
                                 **speaker_argument
                             )
                     if is_audio_data_valid(audio_sentence):
-                        if not isinstance(audio_sentence, torch.Tensor):
-                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported Fairseq wav type: {type(audio_sentence)}"
                             print(error)
                             return False
-                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/tacotron.py b/lib/classes/tts_engines/tacotron.py
index 3285a9b3..0b9a15ca 100644
--- a/lib/classes/tts_engines/tacotron.py
+++ b/lib/classes/tts_engines/tacotron.py
@@ -194,11 +194,16 @@ class Tacotron2(TTSUtils, TTSRegistry, name='tacotron'):
                                 **speaker_argument
                             )
                     if is_audio_data_valid(audio_sentence):
-                        if not isinstance(audio_sentence, torch.Tensor):
-                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported Tacotron2 wav type: {type(audio_sentence)}"
                             print(error)
                             return False
-                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/vits.py b/lib/classes/tts_engines/vits.py
index 09cd7f79..e902c525 100644
--- a/lib/classes/tts_engines/vits.py
+++ b/lib/classes/tts_engines/vits.py
@@ -179,11 +179,16 @@ class Vits(TTSUtils, TTSRegistry, name='vits'):
                                 **speaker_argument
                             )
                     if is_audio_data_valid(audio_sentence):
-                        if not isinstance(audio_sentence, torch.Tensor):
-                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported Vits wav type: {type(audio_sentence)}"
                             print(error)
                             return False
-                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:
diff --git a/lib/classes/tts_engines/yourtts.py b/lib/classes/tts_engines/yourtts.py
index b049f07f..6caaff59 100644
--- a/lib/classes/tts_engines/yourtts.py
+++ b/lib/classes/tts_engines/yourtts.py
@@ -109,11 +109,16 @@ class YourTTS(TTSUtils, TTSRegistry, name='yourtts'):
                             **speaker_argument
                         )
                     if is_audio_data_valid(audio_sentence):
-                        if not isinstance(audio_sentence, torch.Tensor):
-                            error = 'XTTSv2 inference did not return a torch.Tensor'
+                        if isinstance(audio_sentence, torch.Tensor):
+                            audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
+                        elif isinstance(audio_sentence, np.ndarray):
+                            audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
+                        elif isinstance(audio_sentence, (list, tuple)):
+                            audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
+                        else:
+                            error = f"Unsupported YourTTS wav type: {type(audio_sentence)}"
                             print(error)
                             return False
-                        audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         if sentence[-1].isalnum() or sentence[-1] == '—':
                             audio_tensor = trim_audio(audio_tensor.squeeze(), self.params['samplerate'], 0.001, trim_audio_buffer).unsqueeze(0)
                         if audio_tensor is not None and audio_tensor.numel() > 0:

From e06ab173b510f92f3bc6d344b174f70534ba7567 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 12:15:37 -0800
Subject: [PATCH 08/12] ...

---
 lib/classes/tts_engines/bark.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/lib/classes/tts_engines/bark.py b/lib/classes/tts_engines/bark.py
index a149c88f..38ab0074 100644
--- a/lib/classes/tts_engines/bark.py
+++ b/lib/classes/tts_engines/bark.py
@@ -197,7 +197,6 @@ class Bark(TTSUtils, TTSRegistry, name='bark'):
                     #if is_audio_data_valid(audio_sentence):
                     #    audio_sentence = audio_sentence.tolist()
                     if is_audio_data_valid(audio_sentence):
-                        if not isinstance(audio_sentence, torch.Tensor):
                         if isinstance(audio_sentence, torch.Tensor):
                             audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
                         elif isinstance(audio_sentence, np.ndarray):

From f51a6516ff2aa9586408a6dd62e60ff92ddf6029 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 12:51:43 -0800
Subject: [PATCH 09/12] ...

---
 lib/classes/tts_engines/bark.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/lib/classes/tts_engines/bark.py b/lib/classes/tts_engines/bark.py
index 38ab0074..b8c67b89 100644
--- a/lib/classes/tts_engines/bark.py
+++ b/lib/classes/tts_engines/bark.py
@@ -165,6 +165,7 @@ class Bark(TTSUtils, TTSRegistry, name='bark'):
                         """
                     pth_voice_dir = os.path.join(bark_dir, speaker)
                     pth_voice_file = os.path.join(bark_dir, speaker, f'{speaker}.pth')
+                    self.engine.synthesizer.voice_dir = pth_voice_dir
                     tts_dyn_params = {}
                     if not os.path.exists(pth_voice_file) or speaker not in self.engine.speakers:
                         tts_dyn_params['speaker_wav'] = self.params['voice_path']

From 3e19df406669686f264829a93c5d869df886f1f2 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 14:01:59 -0800
Subject: [PATCH 10/12] ...

---
 lib/conf.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/lib/conf.py b/lib/conf.py
index 3d8cf4d7..bc7fa93f 100644
--- a/lib/conf.py
+++ b/lib/conf.py
@@ -45,6 +45,8 @@ os.environ['MPLCONFIGDIR'] = f'{models_dir}/matplotlib'
 os.environ['TESSDATA_PREFIX'] = f'{models_dir}/tessdata'
 os.environ['STANZA_RESOURCES_DIR'] = os.path.join(models_dir, 'stanza')
 os.environ['ARGOS_TRANSLATE_PACKAGE_PATH'] = os.path.join(models_dir, 'argostranslate')
+os.environ['MallocStackLogging'] = 0
+os.environ['MallocStackLoggingNoCompact'] = 0
 os.environ['TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD'] = '1'
 os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1'
 os.environ['PYTORCH_NO_CUDA_MEMORY_CACHING'] = '1'

From e7cd966e31e20d54b24ae77114e1247bf94e0598 Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 15:28:24 -0800
Subject: [PATCH 11/12] ...

---
 lib/conf.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/lib/conf.py b/lib/conf.py
index bc7fa93f..f92a9555 100644
--- a/lib/conf.py
+++ b/lib/conf.py
@@ -52,9 +52,9 @@ os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1'
 os.environ['PYTORCH_NO_CUDA_MEMORY_CACHING'] = '1'
 os.environ['TORCH_CUDA_ENABLE_CUDA_GRAPH'] = '0'
 os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:128,garbage_collection_threshold:0.6,expandable_segments:True'
-os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
-os.environ["CUDA_LAUNCH_BLOCKING"] = "1"
-os.environ["CUDA_CACHE_MAXSIZE"] = "2147483648"
+os.environ['CUDA_DEVICE_ORDER'] = 'PCI_BUS_ID'
+os.environ['CUDA_LAUNCH_BLOCKING'] = '1'
+os.environ['CUDA_CACHE_MAXSIZE'] = '2147483648'
 os.environ['SUNO_OFFLOAD_CPU'] = 'False'
 os.environ['SUNO_USE_SMALL_MODELS'] = 'False'
 if platform.system() == 'Windows':

From ca9f4073b6ba12f712b12e12873c3baf4fc20f6f Mon Sep 17 00:00:00 2001
From: unknown <admin@radsl.net>
Date: Thu, 25 Dec 2025 15:36:54 -0800
Subject: [PATCH 12/12] ...

---
 lib/conf.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/conf.py b/lib/conf.py
index f92a9555..e61c5a7f 100644
--- a/lib/conf.py
+++ b/lib/conf.py
@@ -45,8 +45,8 @@ os.environ['MPLCONFIGDIR'] = f'{models_dir}/matplotlib'
 os.environ['TESSDATA_PREFIX'] = f'{models_dir}/tessdata'
 os.environ['STANZA_RESOURCES_DIR'] = os.path.join(models_dir, 'stanza')
 os.environ['ARGOS_TRANSLATE_PACKAGE_PATH'] = os.path.join(models_dir, 'argostranslate')
-os.environ['MallocStackLogging'] = 0
-os.environ['MallocStackLoggingNoCompact'] = 0
+os.environ['MallocStackLogging'] = '0'
+os.environ['MallocStackLoggingNoCompact'] = '0'
 os.environ['TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD'] = '1'
 os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1'
 os.environ['PYTORCH_NO_CUDA_MEMORY_CACHING'] = '1'