Tortoise new params

requirements.txt

@@ -1,9 +1,10 @@
 # pip install nodejs-bin==16.15.1a4
 ffmpeg-python # Apache 2.0
-gradio==3.33.1
+gradio==3.35.2
 python-dotenv==1.0.0
 soundfile==0.12.1 # torchaudio platform windows
 # run_cmd("pip install sox") # torchaudio platform linux
 # audiocraft==0.0.1 # MIT License
 audiocraft @ git+https://[email protected]/facebookresearch/audiocraft#egg=audiocraft@c9179f8 # MIT License
 vocos==0.0.2 # MIT License
+tortoise @ git+https://github.com/rsxdalv/tortoise-tts#egg=tortoise@f04f95e30245c85996be3af068cba4890952b1a1 # Apache 2.0

server.py

@@ -76,6 +76,9 @@ def reload_config_and_restart_ui():
 
             tab_voice_clone(register_use_as_history_button)
         except Exception as e:
+            with gr.Tab("Bark Voice Clone (!)"):
+                gr.Markdown("Failed to load voice clone demo")
+                gr.Markdown(f"Error: {e}")
             print("Failed to load voice clone demo")
             print(e)
 

src/bark/generation_tab_bark.py

@@ -505,12 +505,14 @@ def generation_tab_bark(tabs):
                 value=None,
                 allow_custom_value=True,
                 visible=False,
+                container=False,
             )
-            old_generation_dropdown.style(container=False)
             copy_old_generation_button = gr.Button(
-                "save", visible=False, elem_classes="btn-sm material-symbols-outlined"
+                "save",
+                visible=False,
+                elem_classes="btn-sm material-symbols-outlined",
+                size="sm",
             )
-            copy_old_generation_button.style(size="sm")
             copy_old_generation_button.click(
                 fn=lambda x: [
                     shutil.copy(x, os.path.join("voices", os.path.basename(x))),
@@ -522,8 +524,8 @@ def generation_tab_bark(tabs):
                 "refresh",
                 visible=False,
                 elem_classes="btn-sm material-symbols-outlined",
+                size="sm",
             )
-            reload_old_generation_dropdown.style(size="sm")
 
             reload_old_generation_dropdown.click(
                 fn=lambda: gr.Dropdown.update(choices=get_npz_files()),

src/bark/setup_seed_ui_bark.py

@@ -4,20 +4,17 @@
 def setup_seed_ui_bark():
     gr.Markdown("Seed")
     with gr.Row():
-        seed_input = gr.Textbox(value="-1", show_label=False)
-        seed_input.style(container=False)
+        seed_input = gr.Textbox(value="-1", show_label=False, container=False)
         set_random_seed_button = gr.Button(
-            "backspace", elem_classes="btn-sm material-symbols-outlined"
+            "backspace", elem_classes="btn-sm material-symbols-outlined", size="sm"
         )
 
-        set_random_seed_button.style(size="sm")
         set_random_seed_button.click(
             fn=lambda: gr.Textbox.update(value="-1"), outputs=[seed_input]
         )
 
         set_old_seed_button = gr.Button(
-            "repeat", elem_classes="btn-sm material-symbols-outlined"
+            "repeat", elem_classes="btn-sm material-symbols-outlined", size="sm"
         )
 
-        set_old_seed_button.style(size="sm")
     return seed_input, set_old_seed_button

src/extensions_loader/ext_callback_save_generation.py

@@ -63,12 +63,15 @@ def ext_callback_save_generation(
     metadata: Dict[str, Any],
 ) -> None:
     for callback in callbacks_save_generation:
-        callback(
-            full_generation=full_generation,
-            audio_array=audio_array,
-            files=files,
-            metadata=metadata,
-        )
+        try:
+            callback(
+                full_generation=full_generation,
+                audio_array=audio_array,
+                files=files,
+                metadata=metadata,
+            )
+        except Exception as e:
+            print("Error in callback_save_generation extension:", e)
 
 
 def ext_callback_save_generation_musicgen(
@@ -78,12 +81,15 @@ def ext_callback_save_generation_musicgen(
     SAMPLE_RATE: int,
 ) -> None:
     for callback in callbacks_save_generation_musicgen:
-        callback(
-            audio_array=audio_array,
-            files=files,
-            metadata=metadata,
-            SAMPLE_RATE=SAMPLE_RATE,
-        )
+        try:
+            callback(
+                audio_array=audio_array,
+                files=files,
+                metadata=metadata,
+                SAMPLE_RATE=SAMPLE_RATE,
+            )
+        except Exception as e:
+            print("Error in callback_save_generation_musicgen extension:", e)
 
 
 if __name__ == "__main__":

src/extensions_loader/extensions/callback_save_generation_ffmpeg.py

@@ -10,6 +10,13 @@
 import base64
 
 
+def check_ffmpeg():
+    if not hasattr(ffmpeg, "input"):
+        raise ImportError(
+            """Incorrect ffmpeg version. Please install ffmpeg-python with `pip install ffmpeg-python`"""
+        )
+
+
 def ndarray_to_base64(arr):
     # Convert ndarray to bytes
     arr_bytes = arr.tobytes()
@@ -31,6 +38,7 @@ def callback_save_generation(
     files: Dict[str, str],
     metadata: Dict[str, Any],
 ) -> None:
+    check_ffmpeg()
     print("Saving generation to", files.get("ogg"))
 
     attach_generation_meta(full_generation, "semantic_prompt", metadata)

src/extensions_loader/extensions/callback_save_generation_musicgen_ffmpeg.py

@@ -6,12 +6,20 @@
 import ffmpeg
 
 
+def check_ffmpeg():
+    if not hasattr(ffmpeg, "input"):
+        raise ImportError(
+            """Incorrect ffmpeg version. Please install ffmpeg-python with `pip install ffmpeg-python`"""
+        )
+
+
 def callback_save_generation_musicgen(
     audio_array: np.ndarray,
     files: Dict[str, str],
     metadata: Dict[str, Any],
     SAMPLE_RATE: int,
 ) -> None:
+    check_ffmpeg()
     print("Saving generation to", files.get("ogg"))
 
     filename = files.get("ogg")

src/history_tab/main.py

@@ -78,8 +78,7 @@ def history_content(
         create_collection_ui(collections_directories_atom)
 
     with gr.Accordion("Gallery Selector (Click to Open)", open=False):
-        history_list_as_gallery = gr.Gallery(value=[])
-        history_list_as_gallery.style(columns=8, object_fit="contain", height="auto")
+        history_list_as_gallery = gr.Gallery(value=[], columns=8, object_fit="contain", height="auto")
     with gr.Row():
         with gr.Column():
             with gr.Row():

src/musicgen/setup_seed_ui_musicgen.py

@@ -4,23 +4,19 @@
 def setup_seed_ui_musicgen():
     gr.Markdown("Seed")
     with gr.Row():
-        seed_input = gr.Number(value=-1, show_label=False)
-        seed_input.style(container=False)
+        seed_input = gr.Number(value=-1, show_label=False, container=False)
         set_random_seed_button = gr.Button(
-            "backspace", elem_classes="btn-sm material-symbols-outlined"
+            "backspace", elem_classes="btn-sm material-symbols-outlined", size="sm"
         )
 
-        set_random_seed_button.style(size="sm")
         set_random_seed_button.click(
             fn=lambda: gr.Number.update(value=-1), outputs=[seed_input]
         )
 
         set_old_seed_button = gr.Button(
-            "repeat", elem_classes="btn-sm material-symbols-outlined"
+            "repeat", elem_classes="btn-sm material-symbols-outlined", size="sm"
         )
 
-        set_old_seed_button.style(size="sm")
-
         def link_seed_cache(seed_cache):
             set_old_seed_button.click(
                 fn=lambda x: gr.Number.update(value=x),