fixes and information updates (#509)

rsxdalv · web-flow · commit 0ad44da3a91e · 2025-05-27T00:02:13.000+03:00
* derank extension_bark_legacy
* attempt install all by default
* Add fixes for decorators to work with non-'text' inputs.
* Clean up .env generator and remove the Bark environment variables from settings.
* Add Audio book extension definitions for future use (extensions not available yet).
* update readme
diff --git a/.gitignore b/.gitignore
@@ -43,6 +43,8 @@ extensions.external.json
 /data/models/kimi-audio/
 /data/models/piper/
 /data/models/ace_step/
+/data/models/bark/
+
 /data/api/presets/
 
 # Ignore temporary files
diff --git a/README.md b/README.md
@@ -20,6 +20,7 @@
   [![Discord](https://img.shields.io/discord/1258772280071295018?label=discord&logo=discord&logoColor=white)](https://discord.gg/V8BKTVRtJ9)
   [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/rsxdalv/tts-webui/blob/main/notebooks/google_colab.ipynb)
   [![GitHub forks](https://img.shields.io/github/forks/rsxdalv/tts-webui?style=social)](https://github.com/rsxdalv/tts-webui/network/members)
+  [![YouTube](https://img.shields.io/badge/YouTube-%23FF0000.svg?logo=YouTube&logoColor=white)](https://www.youtube.com/@TTS-WebUI)
 
 </div>
 
@@ -29,7 +30,7 @@
 
 </div>
 
-| [![Watch the video](https://img.youtube.com/vi/Y8J717tr9t0/sddefault.jpg)](https://youtu.be/Y8J717tr9t0) | [![Watch the video](https://img.youtube.com/vi/ScN2ypewABc/sddefault.jpg)](https://youtu.be/ScN2ypewABc) | [![Watch the video](https://img.youtube.com/vi/JXojhFjZ39k/sddefault.jpg)](https://youtu.be/JXojhFjZ39k) |
+| [![Watch the video](https://img.youtube.com/vi/JXojhFjZ39k/sddefault.jpg)](https://youtu.be/JXojhFjZ39k) | [![Watch the video](https://img.youtube.com/vi/ScN2ypewABc/sddefault.jpg)](https://youtu.be/ScN2ypewABc) | [![Watch the video](https://img.youtube.com/vi/HFtrCnczZQI/sddefault.jpg)](https://youtu.be/HFtrCnczZQI) |
 | :------------------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------: |
 
 <div align="center">
@@ -85,6 +86,14 @@
 
 ## Changelog
 
+May 26:
+* Add fixes for decorators to work with non-'text' inputs.
+* Clean up .env generator and remove the Bark environment variables from settings.
+* Add Audio book extension definitions for future use (extensions not available yet).
+* Fix SeamlessM4T Audio to Audio tab.
+* Update ACE-Step extension.
+* Improve Kokoro TTS API.
+
 May 14:
 * Prepare for Python 3.11 and 3.12 support.
 
diff --git a/extensions.json b/extensions.json
@@ -112,22 +112,6 @@
             "extension_website": "https://github.com/rsxdalv/extension_bark",
             "extension_platform_version": "0.0.1"
         },
-        {
-            "package_name": "extension_bark_voice_clone",
-            "recommended": true,
-            "name": "Bark Voice Clone",
-            "version": "0.0.1",
-            "requirements": "git+https://github.com/rsxdalv/extension_bark_voice_clone@main",
-            "description": "Bark Voice Clone allows cloning voices for use with Bark TTS",
-            "extension_type": "interface",
-            "extension_class": "text-to-speech",
-            "author": "GitMylo",
-            "extension_author": "rsxdalv",
-            "license": "MIT",
-            "website": "https://github.com/gitmylo/bark-voice-cloning-HuBERT-quantizer",
-            "extension_website": "https://github.com/rsxdalv/extension_bark_voice_clone",
-            "extension_platform_version": "0.0.1"
-        },
         {
             "package_name": "extension_stable_audio",
             "recommended": true,
@@ -268,21 +252,6 @@
             "extension_website": "https://github.com/rsxdalv/extension_riffusion",
             "extension_platform_version": "0.0.1"
         },
-        {
-            "package_name": "extension_bark_legacy",
-            "name": "Bark Legacy",
-            "version": "0.0.1",
-            "requirements": "git+https://github.com/rsxdalv/extension_bark_legacy@main",
-            "description": "This is the legacy UI of Bark from TTS-WebUI",
-            "extension_type": "interface",
-            "extension_class": "text-to-speech",
-            "author": "rsxdalv",
-            "extension_author": "rsxdalv",
-            "license": "MIT",
-            "website": "https://github.com/rsxdalv/extension_bark_legacy",
-            "extension_website": "https://github.com/rsxdalv/extension_bark_legacy",
-            "extension_platform_version": "0.0.1"
-        },
         {
             "package_name": "extension_audiocraft_mac",
             "name": "MusicGen (Mac)",
@@ -672,6 +641,82 @@
             "website": "https://github.com/rsxdalv/extension_rvc_training",
             "extension_website": "https://github.com/rsxdalv/extension_rvc_training",
             "extension_platform_version": "0.0.1"
+        },
+        {
+            "package_name": "extension_bark_legacy",
+            "name": "Bark Legacy",
+            "version": "0.0.1",
+            "requirements": "git+https://github.com/rsxdalv/extension_bark_legacy@main",
+            "description": "This is the legacy UI of Bark from TTS-WebUI",
+            "extension_type": "interface",
+            "extension_class": "text-to-speech",
+            "author": "rsxdalv",
+            "extension_author": "rsxdalv",
+            "license": "MIT",
+            "website": "https://github.com/rsxdalv/extension_bark_legacy",
+            "extension_website": "https://github.com/rsxdalv/extension_bark_legacy",
+            "extension_platform_version": "0.0.1"
+        },
+        {
+            "package_name": "extension_bark_voice_clone",
+            "recommended": true,
+            "name": "Bark Voice Clone",
+            "version": "0.0.1",
+            "requirements": "git+https://github.com/rsxdalv/extension_bark_voice_clone@main",
+            "description": "Bark Voice Clone allows cloning voices for use with Bark TTS",
+            "extension_type": "interface",
+            "extension_class": "tools",
+            "author": "GitMylo",
+            "extension_author": "rsxdalv",
+            "license": "MIT",
+            "website": "https://github.com/gitmylo/bark-voice-cloning-HuBERT-quantizer",
+            "extension_website": "https://github.com/rsxdalv/extension_bark_voice_clone",
+            "extension_platform_version": "0.0.1"
+        },
+        {
+            "package_name": "extension_ebook2audiobook",
+            "name": "Ebook2Audiobook (Not available yet)",
+            "version": "0.0.1",
+            "requirements": "git+https://github.com/rsxdalv/extension_ebook2audiobook@main",
+            "description": "Ebook2Audiobook allows converting ebooks to audiobooks",
+            "extension_type": "interface",
+            "extension_class": "tools",
+            "author": "rsxdalv",
+            "extension_author": "rsxdalv",
+            "license": "MIT",
+            "website": "https://github.com/rsxdalv/extension_ebook2audiobook",
+            "extension_website": "https://github.com/rsxdalv/extension_ebook2audiobook",
+            "extension_platform_version": "0.0.1"
+        },
+        {
+            "package_name": "extension_epub2tts",
+            "name": "EPub2TTS (Not available yet)",
+            "version": "0.0.1",
+            "requirements": "git+https://github.com/rsxdalv/extension_epub2tts@main",
+            "description": "EPub2TTS allows converting ebooks to audiobooks",
+            "extension_type": "interface",
+            "extension_class": "tools",
+            "author": "rsxdalv",
+            "extension_author": "rsxdalv",
+            "license": "MIT",
+            "website": "https://github.com/rsxdalv/extension_epub2tts",
+            "extension_website": "https://github.com/rsxdalv/extension_epub2tts",
+            "extension_platform_version": "0.0.1"
+        },
+        {
+            "package_name": "extension_audiobook_generator",
+            "name": "Audiobook Generator (Not available yet)",
+            "version": "0.0.1",
+            "requirements": "git+https://github.com/rsxdalv/extension_audiobook_generator@main",
+            "description": "Audiobook Generator allows converting ebooks to audiobooks",
+            "extension_type": "interface",
+            "extension_class": "tools",
+            "author": "rsxdalv",
+            "extension_author": "rsxdalv",
+            "license": "MIT",
+            "website": "https://github.com/rsxdalv/extension_audiobook_generator",
+            "extension_website": "https://github.com/rsxdalv/extension_audiobook_generator",
+            "extension_platform_version": "0.0.1"
         }
     ],
     "decorators": [
diff --git a/extensions/builtin/extension_decorator_save_ffmpeg/main.py b/extensions/builtin/extension_decorator_save_ffmpeg/main.py
@@ -113,6 +113,7 @@ def callback_save_generation_musicgen(
     print("Saving generation to", filename)
 
     input_data = audio_array.tobytes()
+    metadata["text"] = metadata.get("text", "")
     metadata["text"] = _double_escape_quotes(metadata["text"])
     metadata["text"] = _double_escape_newlines(metadata["text"])
     metadata_str = json.dumps(metadata, ensure_ascii=False)
@@ -216,6 +217,7 @@ def _attach_generation_meta(full_generation, arg1, metadata):
 
     _attach_generation_meta(full_generation, "semantic_prompt", metadata)
     _attach_generation_meta(full_generation, "coarse_prompt", metadata)
+    metadata["text"] = metadata.get("text", "")
     metadata["text"] = _double_escape_quotes(metadata["text"])
     metadata["text"] = _double_escape_newlines(metadata["text"])
 
diff --git a/installer_scripts/js/initializeApp.js b/installer_scripts/js/initializeApp.js
@@ -157,11 +157,12 @@ async function pip_install_all(fi = false) {
       "Dependencies are already up to date, skipping pip installs..."
     );
 
-  const pip_install_all_choice = await menu(
-    ["Yes", "No"],
-    `Attempt single pip install of all dependencies (potentially faster)?
-    (use arrow keys to move, enter to select)`
-  );
+  // const pip_install_all_choice = await menu(
+  //   ["Yes", "No"],
+  //   `Attempt single pip install of all dependencies (potentially faster)?
+  //   (use arrow keys to move, enter to select)`
+  // );
+  const pip_install_all_choice = "Yes";
 
   if (pip_install_all_choice === "Yes") {
     try {
diff --git a/react-ui/src/pages/text-to-speech/bark/bark_settings.tsx b/react-ui/src/pages/text-to-speech/bark/bark_settings.tsx
@@ -83,29 +83,6 @@ const BarkSettingsPage = () => {
         console.error("Error:", error);
       }
     }
-
-    const env_keys = [
-      "env_suno_use_small_models",
-      "env_suno_enable_mps",
-      "env_suno_offload_cpu",
-    ];
-    const isEnv = env_keys.includes(name);
-
-    if (isEnv) {
-      try {
-        await fetch("/api/gradio/save_env_variables_bark", {
-          method: "POST",
-          body: JSON.stringify(
-            Object.fromEntries(
-              env_keys.map((key) => [key, newBarkSettingsParams[key]])
-            )
-          ),
-        });
-        setBarkSettingsParams(newBarkSettingsParams);
-      } catch (error) {
-        console.error("Error:", error);
-      }
-    }
   };
 
   return (
@@ -177,39 +154,10 @@ const BarkSettingsPage = () => {
             />
           </div>
 
-          <div className="flex flex-col gap-y-2 cell">
-            <div className="flex gap-x-2 items-center">
-              <label className="text-md">Environment (requires restart):</label>
-            </div>
-            <SwitchWithLabel
-              label="Use small models"
-              name="env_suno_use_small_models"
-              value={barkSettingsParams.env_suno_use_small_models}
-              onChange={handleChange}
-            />
-            <SwitchWithLabel
-              label="Enable MPS"
-              name="env_suno_enable_mps"
-              value={barkSettingsParams.env_suno_enable_mps}
-              onChange={handleChange}
-            />
-            <SwitchWithLabel
-              label="Offload GPU models to CPU"
-              name="env_suno_offload_cpu"
-              value={barkSettingsParams.env_suno_offload_cpu}
-              onChange={handleChange}
-            />
-          </div>
-
           <Button
             variant="outline"
             onClick={() => {
-              setBarkSettingsParams({
-                ...barkSettingsParams,
-                env_suno_use_small_models: false,
-                env_suno_enable_mps: false,
-                env_suno_offload_cpu: false,
-              });
+              setBarkSettingsParams({ ...barkSettingsParams });
             }}
           >
             Reset to defaults
diff --git a/react-ui/src/tabs/BarkSettingsParams.tsx b/react-ui/src/tabs/BarkSettingsParams.tsx
@@ -4,10 +4,6 @@ import router from "next/router";
 export const barkSettingsId = "barkSettingsParams.v4";
 
 export type BarkSettingsParams = {
-  env_suno_use_small_models: boolean;
-  env_suno_enable_mps: boolean;
-  env_suno_offload_cpu: boolean;
-
   text_use_gpu: boolean;
   text_use_small: boolean;
   coarse_use_gpu: boolean;
@@ -19,10 +15,6 @@ export type BarkSettingsParams = {
 };
 
 export const initialBarkSettingsParams: BarkSettingsParams = {
-  env_suno_use_small_models: false,
-  env_suno_enable_mps: false,
-  env_suno_offload_cpu: false,
-
   text_use_gpu: false,
   text_use_small: false,
   coarse_use_gpu: false,
diff --git a/tts_webui/decorators/__init__.py b/tts_webui/decorators/__init__.py
@@ -1,10 +1,10 @@
-from .decorator_add_base_filename import decorator_add_base_filename
-from .decorator_add_date import decorator_add_date
-from .decorator_add_model_type import decorator_add_model_type
-from .decorator_apply_torch_seed import decorator_apply_torch_seed
-from .decorator_log_generation import decorator_log_generation
+from .decorator_add_base_filename import decorator_add_base_filename, decorator_add_base_filename_generator, format_date_for_file
+from .decorator_add_date import decorator_add_date, decorator_add_date_generator
+from .decorator_add_model_type import decorator_add_model_type, decorator_add_model_type_generator
+from .decorator_apply_torch_seed import decorator_apply_torch_seed, decorator_apply_torch_seed_generator
+from .decorator_log_generation import decorator_log_generation, decorator_log_generation_generator
 from .decorator_save_metadata import decorator_save_metadata
 from .decorator_save_musicgen_npz import decorator_save_musicgen_npz
-from .decorator_save_wav import decorator_save_wav
+from .decorator_save_wav import decorator_save_wav, decorator_save_wav_generator
 from .gradio_dict_decorator import dictionarize
-from .log_function_time import log_function_time
+from .log_function_time import log_function_time, log_generator_time
diff --git a/tts_webui/decorators/decorator_add_base_filename.py b/tts_webui/decorators/decorator_add_base_filename.py
@@ -14,7 +14,7 @@ def format_date_for_file(date: datetime):
 
 
 def _create_base_filename(kwargs, result_dict):
-    prompt = kwargs["text"]
+    prompt = kwargs.get("text", "")
     is_long = result_dict.get("long", False)
     base_filename = format_filename(
         title=prompt_to_title(prompt) + ("_long" if is_long else ""),
diff --git a/tts_webui/utils/log_generation.py b/tts_webui/utils/log_generation.py
@@ -11,7 +11,18 @@ def custom_repr(value):
 
 
 def StringifyParams(x):
-    params = ",\n    ".join(f"{k}={custom_repr(v)}" for k, v in x.items())
+    def filter_keys(k):
+        if k == "outputs":
+            return False
+        if k == "_type":
+            return False
+        if k == "text":
+            return False
+        return True
+
+    params = ",\n    ".join(
+        f"{k}={custom_repr(v)}" for k, v in x.items() if filter_keys(k)
+    )
     return f"{_get_typed_dict_name(x)}(\n    {params}\n)"
 
 
diff --git a/tts_webui/utils/setup_or_recover.py b/tts_webui/utils/setup_or_recover.py
