Additional note in finetune re multiple GPU support

Dev

.gitignore

@@ -1,6 +1,7 @@
 voices/*.*
 models/*.*
 outputs/*.*
+finetune/*.*
 config.json
 confignew.json
 models.json

README.md

@@ -25,10 +25,9 @@ AllTalk is an updated version of the Coqui_tts extension for Text Generation web
 - 🟠 [API Suite and JSON-CURL](https://github.com/erew123/alltalk_tts?#-api-suite-and-json-curl)
 - 🔴 [Future to-do list & Upcoming updates](https://github.com/erew123/alltalk_tts?#-future-to-do-list)
 
-#### Updates
+## 🔄 Updates
-Dec 25th - Applied a small update to avoid a possible race condition on file naming with small sentences when generating narrator/character speech. Please update if you are experiencing missing small bits of missing audio on narrator/character.
+Please check the below link to find a list of all recent updates and changes.
-
+####     🔄 **Updates list & bug fixes list** can be found [here](https://github.com/erew123/alltalk_tts/issues/25)
-####     🔄 **Minor updates/bug fixes list** can be found [here](https://github.com/erew123/alltalk_tts/issues/25)
 
 ## 🟩 Installation on Text generation web UI
 This has been tested on the current Dec 2023 release of Text generation webUI. If you have not updated it for a while, you may wish to update Text generation webUI, [instructions here](https://github.com/oobabooga/text-generation-webui?tab=readme-ov-file#how-to-install)
@@ -102,6 +101,7 @@ I did leave a mistake in the `/extensions/alltalk_tts/.gitignore` file at one po
 voices/*.*
 models/*.*
 outputs/*.*
+finetune/*.*
 config.json
 confignew.json
 models.json
@@ -254,10 +254,29 @@ I would suggest following [Problems Updating](https://github.com/erew123/alltalk
 As far as I am aware, these are to do with the chrome browser the gradio text-generation-webui in some way. I raised an issue about this on the text-generation-webui [here](https://github.com/oobabooga/text-generation-webui/issues/4788) where you can see that AllTalk is not loaded and the messages persist. Either way, this is more a warning than an actual issue, so shouldnt affect any functionality of either AllTalk or text-generation-webui, they are more just an annoyance.
 </details>
 
+#### 🟨 I have multiple GPU's and I have problems running Finetuning
+
+<details>
+	<summary>Click to expand</summary><br>
+	
+Finetuning pulls in various other scripts and some of those scripts can have issues with multiple Nvidia GPU's being present. Until the people that created those other scripts fix up their code, there is a workaround to temporarily tell your system to only use the 1x of your Nvidia GPU's. To do this:
+
+- **Windows** - You will start the script with `set CUDA_VISIBLE_DEVICES=0 && python finetune.py`<br>
+After you have completed training, you can reset back with `set CUDA_VISIBLE_DEVICES=`<br>
+   
+- **Linux** - You will start the script with `CUDA_VISIBLE_DEVICES=0 python finetune.py`<br>
+After you have completed training, you can reset back with `unset CUDA_VISIBLE_DEVICES`<br>
+
+Rebooting your system will also unset this. The setting is only applied temporarily.
+
+Depending on which of your Nvidia GPU's is the more powerful one, you can change the `0` to `1` or whichever of your GPU's is the most powerful.
+
+</details>
+
 ## ⚫ Finetuning a model
 If you have a voice that the model doesnt quite reproduce correctly, or indeed you just want to improve the reproduced voice, then finetuning is a way to train your "XTTSv2 local" model **(stored in `/alltalk_tts/models/xxxxx/`)** on a specific voice. For this you will need:
 
-- An Nvidia graphics card
+- An Nvidia graphics card. (Please see this [note](https://github.com/erew123/alltalk_tts#i-have-multiple-gpus-and-i-have-problems-running-finetuning) if you have multiple Nvidia GPU's).
 - To install a few portions of the Nvidia CUDA 11.8 Toolkit (this will not impact text-generation-webui's cuda setup.
 - 18GB of disk space free (most of this is used temporarily)
 - At least 2 minutes of good quality speech from your chosen speaker in mp3, wav or flacc format, in one or more files (have tested as far as 20 minutes worth of audio).
@@ -315,6 +334,18 @@ As mentioned you must have a small portion of the Nvidia CUDA Toolkit **11.8** i
 11) Follow the on-screen instructions when the web interface starts up.
 12) When you have finished finetuning, the final tab will tell you what to do with your files and how to move your newly trained model to the correct location on disk.
 
+#### ⚫ Using a Finetuned model in Text-generation-webui
+
+At the end of the finetune process, you will have an option to `Compact and move model to /trainedmodel/` this will compact the raw training file and move it to `/model/trainedmodel/`. When AllTalk starts up within Text-generation-webui, if it finds a model in this location a new loader will appear in the interface for `XTTSv2 FT` and you can use this to load your finetuned model. <br><br>**Be careful** not to train a new model from the base model, then overwrite your current `/model/trainedmodel/` **if** you want a seperately trained model. This is why there is an `OPTION B` to move your just trained model to `/models/lastfinetuned/`.
+
+#### ⚫ Training one model with multiple voices
+
+At the end of the finetune process, you will have an option to `Compact and move model to /trainedmodel/` this will compact the raw training file and move it to `/model/trainedmodel/`. This model will become available when you start up finetuning. You will have a choice to train the Base Model or the `Existing finetuned model` (which is the one in `/model/trainedmodel/`). So you can use this to keep further training this model with additional voices, then copying it back to `/model/trainedmodel/` at the end of training.
+
+#### ⚫ Do I need to keep the raw training data/model?
+
+If you've compacted and moved your model, its highly unlikely you would want to keep that data, however the choice is there to keep it if you wish. It will be between 5-10GB in size, so most people will want to delete it.
+
 ## 🔵🟢🟡 DeepSpeed Installation Options
 **NOTE**: You **DO NOT** need to set Text-generation-webUI's **--deepspeed** setting for AllTalk to be able to use DeepSpeed. These are two completely separate things and incorrectly setting that on Text-generation-webUI may cause other complications.
 
@@ -471,7 +502,7 @@ def get_shm_size():
 ## 🟦 Running AllTalk as a standalone app
 AllTalk will run as a standalone app, as long as you install its requirements files into whatever Python environment you are using. You can follow the steps to install the AllTalk's requirements into whatever Python environment you wish. Because I dont know what Python environment you are wanting to use, I can only give you a loose set of installation instructions. 
 
-Please note, at time of writing, the TTS engine requires Python **3.9.x** to **3.11.x** [TTS Engine details here](https://pypi.org/project/TTS/)
+Please note, at time of writing, the TTS engine requires Python **3.9.x** to **3.11.x** [TTS Engine details here](https://pypi.org/project/TTS/). AllTalk and its requirements are tested on Python 3.11.x.
 
 #### 🟦 (Option 1) I already have AllTalk installed as an extension of Text-generation-webui
 If you already have AllTalk as a extension of Text-generation-webui, and wish to run it as standalone, load Text-generation-webui's Python environment `cmd_windows.bat`, `./cmd_linux.sh`, `cmd_macos.sh` or `cmd_wsl.bat`, move into the AllTalk folder `cd extensions` > `cd alltalk_tts` and start AllTalk with `python script.py`. There is nothing beyond this you would need to do.
@@ -497,9 +528,71 @@ Deepspeed and other such things can be installed. Please read the relevant instr
 ### 🟠Overview
 The Text-to-Speech (TTS) Generation API allows you to generate speech from text input using various configuration options. This API supports both character and narrator voices, providing flexibility for creating dynamic and engaging audio content.
 
-- URL: `http://127.0.0.1:7851/api/tts-generate`<br>
+#### 🟠 Ready Endpoint<br>
-- Method: `POST`<br>
+Check if the Text-to-Speech (TTS) service is ready to accept requests.
-- Content-Type: `application/x-www-form-urlencoded`<br>
+
+- URL: `http://127.0.0.1:7851/api/ready`<br> - Method: `GET`<br> 
+
+   `curl -X GET "http://127.0.0.1:7851/api/ready"`
+
+  Response: `Ready`
+
+#### 🟠 Voices List Endpoint<br>
+Retrieve a list of available voices for generating speech.
+
+- URL: `http://127.0.0.1:7851/api/voices`<br> - Method: `GET`<br>
+
+   `curl -X GET "http://127.0.0.1:7851/api/voices"`
+
+   JSON return: `{"voices": ["voice1.wav", "voice2.wav", "voice3.wav"]}`
+
+#### 🟠 Preview Voice Endpoint
+Generate a preview of a specified voice with hardcoded settings.
+
+- URL: `http://127.0.0.1:7851/api/previewvoice/`<br> - Method: `POST`<br> - Content-Type: `application/x-www-form-urlencoded`<br>
+
+   `curl -X POST "http://127.0.0.1:7851/api/previewvoice/" -F "voice=female_01.wav"`
+
+   Replace `female_01.wav` with the name of the voice sample you want to hear.
+
+   JSON return: `{"status": "generate-success", "output_file_path": "/path/to/outputs/api_preview_voice.wav", "output_file_url": "http://127.0.0.1:7851/audio/api_preview_voice.wav"}`
+
+#### 🟠 Switching Model Endpoint<br>
+
+- URL: `http://127.0.0.1:7851/api/reload`<br> - Method: `POST`<br><br>
+   `curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=API%20Local"`<br>
+   `curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=API%20TTS"`<br>
+   `curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=XTTSv2%20Local"`<br>
+
+   Switch between the 3 models respectively.
+
+   `curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=XTTSv2%20FT"`<br>
+
+   If you have a finetuned model in `/models/trainedmodel/` (will error otherwise)
+
+   JSON return `{"status": "model-success"}`
+
+#### 🟠 Switch DeepSpeed Endpoint<br>
+
+- URL: `http://127.0.0.1:7851/api/deepspeed`<br> - Method: `POST`<br><br>
+   `curl -X POST "http://127.0.0.1:7851/api/deepspeed?new_deepspeed_value=True"`
+
+   Replace True with False to disable DeepSpeed mode.
+
+   JSON return `{"status": "deepspeed-success"}`
+
+#### 🟠 Switching Low VRAM Endpoint<br>
+
+- URL: `http://127.0.0.1:7851/api/lowvramsetting`<br> - Method: `POST`<br><br>
+   `curl -X POST "http://127.0.0.1:7851/api/lowvramsetting?new_low_vram_value=True"`
+
+   Replace True with False to disable Low VRAM mode.
+
+   JSON return `{"status": "lowvram-success"}`
+
+### 🟠 TTS Generation Endpoint
+
+- URL: `http://127.0.0.1:7851/api/tts-generate`<br> - Method: `POST`<br> - Content-Type: `application/x-www-form-urlencoded`<br>
 
 ### 🟠 Example command lines
 Standard TTS speech Example (standard text) generating a time-stamped file<br>
@@ -556,22 +649,22 @@ Example:
 
 🟠 **language**: Choose the language for TTS. Options:
 
-`ar Arabic`<br>
+`ar` Arabic<br>
-`zh-cn Chinese (Simplified)`<br>
+`zh-cn` Chinese (Simplified)<br>
-`cs Czech`<br>
+`cs` Czech<br>
-`nl Dutch`<br>
+`nl` Dutch<br>
-`en English`<br>
+`en` English<br>
-`fr French`<br>
+`fr` French<br>
-`de German`<br>
+`de` German<br>
-`hu Hungarian`<br>
+`hu` Hungarian<br>
-`it Italian`<br>
+`it` Italian<br>
-`ja Japanese`<br>
+`ja` Japanese<br>
-`ko Korean`<br>
+`ko` Korean<br>
-`pl Polish`<br>
+`pl` Polish<br>
-`pt Portuguese`<br>
+`pt` Portuguese<br>
-`ru Russian`<br>
+`ru` Russian<br>
-`es Spanish`<br>
+`es` Spanish<br>
-`tr Turkish`<br>
+`tr` Turkish<br>
 
 `-d "language=en"`<br>
 
@@ -584,12 +677,12 @@ Example:
 `-d "output_file_timestamp=true"`<br>
 `-d "output_file_timestamp=false"`
 
-🟠 **autoplay**: Feature not yet available. Enable or disable autoplay. Still needs to be specified in the JSON request.
+🟠 **autoplay**: Enable or disable playing the generated TTS to your standard sound output device at time of TTS generation.
 
 `-d "autoplay=true"`<br>
 `-d "autoplay=false"`
 
-🟠 **autoplay_volume**: Feature not yet available. Set the autoplay volume. Should be between 0.1 and 1.0. Still needs to be specified in the JSON request.
+🟠 **autoplay_volume**: Set the autoplay volume. Should be between 0.1 and 1.0. Needs to be specified in the JSON request even if autoplay is false.
 
 `-d "autoplay_volume=0.8"`
 
@@ -604,29 +697,6 @@ Example JSON TTS Generation Response:
 
 `{"status": "generate-success", "output_file_path": "C:\text-generation-webui\extensions\alltalk_tts\outputs\myoutputfile_1703149973.wav", "output_file_url": "http://127.0.0.1:7851/audio/myoutputfile_1703149973.wav"}`
 
-🟠 **Switching Model**<br><br>
-`curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=API%20Local"`<br>
-`curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=API%20TTS"`<br>
-`curl -X POST "http://127.0.0.1:7851/api/reload?tts_method=XTTSv2%20Local"`<br>
-
-Switch between the 3 models respectively.
-
-JSON return `{"status": "model-success"}`
-
-🟠 **Switch DeepSpeed**<br><br>
-`curl -X POST "http://127.0.0.1:7851/api/deepspeed?new_deepspeed_value=True"`
-
-Replace True with False to disable DeepSpeed mode.
-
-JSON return `{"status": "deepspeed-success"}`
-
-🟠 **Switching Low VRAM**<br><br>
-`curl -X POST "http://127.0.0.1:7851/api/lowvramsetting?new_low_vram_value=True"`
-
-Replace True with False to disable Low VRAM mode.
-
-JSON return `{"status": "lowvram-success"}`
-
 ### 🔴 Future to-do list
 - Voice output within the command prompt/terminal (TBD).
 - Correct a few spelling mistakes in the documentation.

finetune.py

@@ -48,6 +48,20 @@ else:
     print("[FINETUNE]  \033[91mWarning\033[0m modeldownload.json is missing. Please run this script in the /alltalk_tts/ folder")
     sys.exit(1)
 
+##################################################
+#### Check to see if a finetuned model exists ####
+##################################################
+# Set the path to the directory
+trained_model_directory = this_dir / "models" / "trainedmodel"
+# Check if the directory "trainedmodel" exists
+finetuned_model = trained_model_directory.exists()
+# If the directory exists, check for the existence of the required files
+# If true, this will add a extra option in the Gradio interface for loading Xttsv2 FT
+if finetuned_model:
+    required_files = ["model.pth", "config.json", "vocab.json", "mel_stats.pth", "dvae.pth"]
+    finetuned_model = all((trained_model_directory / file).exists() for file in required_files)
+basemodel_or_finetunedmodel = True
+
 #####################
 #### STEP 1 BITS ####
 #####################
@@ -272,6 +286,8 @@ def format_audio_list(target_language, whisper_model, out_path, gradio_progress=
     # deallocate VRAM and RAM
     del asr_model, final_eval_set, final_training_set, new_data_df, existing_metadata
     gc.collect()
+    existing_train_df = None
+    existing_eval_df = None
     print("[FINETUNE] Train CSV:", train_metadata_path)
     print("[FINETUNE] Eval CSV:", eval_metadata_path)
     print("[FINETUNE] Audio Total:", audio_total_size)
@@ -289,6 +305,13 @@ from TTS.tts.layers.xtts.trainer.gpt_trainer import GPTArgs, GPTTrainer, GPTTrai
 from TTS.utils.manage import ModelManager
 
 
+def basemodel_or_finetunedmodel_choice(value):
+    global basemodel_or_finetunedmodel 
+    if value == "Base Model":
+        basemodel_or_finetunedmodel = True
+    elif value == "Existing finetuned model":
+        basemodel_or_finetunedmodel = False
+
 def train_gpt(language, num_epochs, batch_size, grad_acumm, train_csv, eval_csv, output_path, max_audio_length=255995):
     #  Logging parameters
     RUN_NAME = "XTTS_FT"
@@ -321,12 +344,22 @@ def train_gpt(language, num_epochs, batch_size, grad_acumm, train_csv, eval_csv,
     # Add here the configs of the datasets
     DATASETS_CONFIG_LIST = [config_dataset]
 
-    # XTTS model checkpoints for fine-tuning.
+    if basemodel_or_finetunedmodel:
-    TOKENIZER_FILE = str(this_dir / base_path / model_path / "vocab.json")  # vocab.json file
+        # BASE XTTS model checkpoints for fine-tuning.
-    XTTS_CHECKPOINT = str(this_dir / base_path / model_path / "model.pth")  # model.pth file
+        print("[FINETUNE] Starting finetuning on Base Model")
-    XTTS_CONFIG_FILE = str(this_dir / base_path / model_path / "config.json")  # config.json file
+        TOKENIZER_FILE = str(this_dir / base_path / model_path / "vocab.json")
-    DVAE_CHECKPOINT = str(this_dir / base_path / model_path / "dvae.pth")
+        XTTS_CHECKPOINT = str(this_dir / base_path / model_path / "model.pth")
-    MEL_NORM_FILE = str(this_dir / base_path / model_path / "mel_stats.pth")
+        XTTS_CONFIG_FILE = str(this_dir / base_path / model_path / "config.json")
+        DVAE_CHECKPOINT = str(this_dir / base_path / model_path / "dvae.pth")
+        MEL_NORM_FILE = str(this_dir / base_path / model_path / "mel_stats.pth")
+    else:
+        # FINETUNED XTTS model checkpoints for fine-tuning.
+        print("[FINETUNE] Starting finetuning on Existing Finetuned Model")
+        TOKENIZER_FILE = str(this_dir / base_path / "trainedmodel" / "vocab.json")
+        XTTS_CHECKPOINT = str(this_dir / base_path / "trainedmodel" / "model.pth")
+        XTTS_CONFIG_FILE = str(this_dir / base_path / "trainedmodel" / "config.json")
+        DVAE_CHECKPOINT = str(this_dir / base_path / "trainedmodel" / "dvae.pth")
+        MEL_NORM_FILE = str(this_dir / base_path / "trainedmodel" / "mel_stats.pth")
 
     # init args and config
     model_args = GPTArgs(
@@ -417,9 +450,14 @@ def train_gpt(language, num_epochs, batch_size, grad_acumm, train_csv, eval_csv,
     trainer_out_path = trainer.output_path
 
     # deallocate VRAM and RAM
-    del model, trainer, train_samples, eval_samples
+    del model, trainer, train_samples, eval_samples, config, model_args, config_dataset
     gc.collect()
-
+    train_samples = None
+    eval_samples = None
+    config_dataset = None
+    trainer = None
+    model = None
+    model_args = None
     return XTTS_CONFIG_FILE, XTTS_CHECKPOINT, TOKENIZER_FILE, trainer_out_path, speaker_ref
 
 ##########################
@@ -514,7 +552,7 @@ def compact_model():
 
     # Check if the best model file exists
     if best_model_path_str is None:
-        print("No trained model was found.")
+        print("[FINETUNE] No trained model was found.")
         return "No trained model was found."
 
     # Convert model_path_str to Path
@@ -524,7 +562,7 @@ def compact_model():
     try:
         checkpoint = torch.load(best_model_path, map_location=torch.device("cpu"))
     except Exception as e:
-        print("Error loading checkpoint:", e)
+        print("[FINETUNE] Error loading checkpoint:", e)
         raise
 
     del checkpoint["optimizer"]
@@ -565,20 +603,128 @@ def compact_model():
     return "Model copied to '/models/trainedmodel/'"
 
 
+def compact_lastfinetuned_model():
+    this_dir = Path(__file__).parent.resolve()
+
+    best_model_path_str = find_latest_best_model(this_dir / "finetune" / "tmp-trn" / "training")
+
+    # Check if the best model file exists
+    if best_model_path_str is None:
+        print("[FINETUNE] No trained model was found.")
+        return "No trained model was found."
+
+    # Convert model_path_str to Path
+    best_model_path = Path(best_model_path_str)
+
+    # Attempt to load the model
+    try:
+        checkpoint = torch.load(best_model_path, map_location=torch.device("cpu"))
+    except Exception as e:
+        print("[FINETUNE] Error loading checkpoint:", e)
+        raise
+
+    del checkpoint["optimizer"]
+
+    for key in list(checkpoint["model"].keys()):
+        if "dvae" in key:
+            del checkpoint["model"][key]
+
+    # Define the target directory
+    target_dir = this_dir / "models" / "lastfinetuned"
+
+    # Create the target directory if it doesn't exist
+    target_dir.mkdir(parents=True, exist_ok=True)
+
+    # Save the modified checkpoint in the target directory
+    torch.save(checkpoint, target_dir / "model.pth")
+
+    # Specify the files you want to copy
+    files_to_copy = ["vocab.json", "config.json", "dvae.pth", "mel_stats.pth"]
+
+    for file_name in files_to_copy:
+        src_path = this_dir / base_path / base_model_path / file_name
+        dest_path = target_dir / file_name
+        shutil.copy(str(src_path), str(dest_path))
+
+    source_wavs_dir = this_dir / "finetune" / "tmp-trn" / "wavs"
+    target_wavs_dir = target_dir / "wavs"
+    target_wavs_dir.mkdir(parents=True, exist_ok=True)
+
+    # Iterate through files in the source directory
+    for file_path in source_wavs_dir.iterdir():
+        # Check if it's a file and larger than 1000 KB
+        if file_path.is_file() and file_path.stat().st_size > 1000 * 1024:
+            # Copy the file to the target directory
+            shutil.copy(str(file_path), str(target_wavs_dir / file_path.name))
+    
+    print("[FINETUNE] Model copied to '/models/lastfinetuned/'")
+    return "Model copied to '/models/lastfinetuned/'"
+
+
+def compact_legacy_model():
+    this_dir = Path(__file__).parent.resolve()
+
+    best_model_path_str = os.path.join(this_dir, "finetune", "best_model.pth")
+
+    # Check if the best model file exists
+    if not os.path.exists(best_model_path_str):
+        print("[FINETUNE] No model called best_model.pth was found in /finetune/")
+        return "No model called best_model.pth was found in /finetune/"
+
+    # Convert model_path_str to Path
+    best_model_path = Path(best_model_path_str)
+
+    # Attempt to load the model
+    try:
+        checkpoint = torch.load(best_model_path, map_location=torch.device("cpu"))
+    except Exception as e:
+        print("[FINETUNE] Error loading checkpoint:", e)
+        raise
+
+    del checkpoint["optimizer"]
+
+    for key in list(checkpoint["model"].keys()):
+        if "dvae" in key:
+            del checkpoint["model"][key]
+
+    # Define the target directory
+    target_dir = this_dir / "finetune"
+
+    # Create the target directory if it doesn't exist
+    target_dir.mkdir(parents=True, exist_ok=True)
+
+    # Save the modified checkpoint in the target directory
+    torch.save(checkpoint, target_dir / "model.pth")
+
+    print("[FINETUNE] model.pth created in '/finetune/'")
+    return "model.pth created in '/finetune/'"
+
+
 def delete_training_data():
     # Define the folder to be deleted
     folder_to_delete = Path(this_dir / "finetune" / "tmp-trn")
 
     # Check if the folder exists before deleting
     if folder_to_delete.exists():
-        # Delete the folder and its contents
+        # Iterate over all files and subdirectories
-        shutil.rmtree(folder_to_delete)
+        for item in folder_to_delete.iterdir():
-        print(f"[FINETUNE] Folder {folder_to_delete} and its contents deleted successfully.")
+            # Exclude trainer_0_log.txt from deletion
-        return "Folder '/finetune/tmp-trn/' and its contents deleted successfully."
+            if item.name != "trainer_0_log.txt":
+                try:
+                    if item.is_file():
+                        item.unlink()
+                    elif item.is_dir():
+                        shutil.rmtree(item)
+                except PermissionError:
+                    print(f"[FINETUNE] PermissionError: Could not delete {item}. Skipping.")
+
+        print(f"[FINETUNE] Folder {folder_to_delete} contents (excluding trainer_0_log.txt) deleted successfully.")
+        return "Folder '/finetune/tmp-trn/' contents (excluding trainer_0_log.txt) deleted successfully."
     else:
         print(f"[FINETUNE] Folder {folder_to_delete} does not exist.")
         return "Folder '/finetune/tmp-trn/' does not exist."
 
+
 def delete_voice_sample_contents():
     # Define the folder to be cleared
     folder_to_clear = Path(this_dir / "finetune" / "put-voice-samples-in-here")
@@ -729,20 +875,21 @@ if __name__ == "__main__":
                 f"""
                 ## <u>Finetuning Information</u><br>
                 ### 🟥 <u>Important Note</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟥 <span style="color: #3366ff;">finetune.py</span> needs to be run from the <span style="color: #3366ff;">/alltalk_tts/</span> folder. Don't move the location of this script.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - <span style="color: #3366ff;">finetune.py</span> needs to be run from the <span style="color: #3366ff;">/alltalk_tts/</span> folder. Don't move the location of this script.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟥 Have you run AllTalk at least once? It needs to have downloaded+updated the voice model, before we can finetune it.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Have you run AllTalk at least once? It needs to have downloaded+updated the voice model, before we can finetune it.
                 ### 🟦 <u>What you need to run finetuning</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 An Nvidia GPU. Tested on Windows with extended shared VRAM and training used about 16GB's total (which worked on a 12GB card).
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - An Nvidia GPU. Tested on Windows with extended shared VRAM and training used about 16GB's total (which worked on a 12GB card).
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 I have not been able to test this on a GPU with less than 12GB of VRAM, so cannot say if it will work or how that would affect performance.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - If you have multiple Nvidia GPU's in your system, please see this [important note](https://github.com/erew123/alltalk_tts#-i-have-multiple-gpus-and-i-have-problems-running-finetuning).
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 <span style="color: red;">Version 11.8</span> of Nvidia cuBLAS and cuDNN (guide below). Only 11.8 of cuBLAS and cuDNN work for this process currently.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - I have not been able to test this on a GPU with less than 12GB of VRAM, so cannot say if it will work or how that would affect performance.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Minimum <span style="color: red;">18GB</span> free disk space (most of it is used temporarily).
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - <span style="color: red;">Version 11.8</span> of Nvidia cuBLAS and cuDNN (guide below). Only 11.8 of cuBLAS and cuDNN work for this process currently.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Some decent quality audio, multiple files if you like. Minimum of 2 minutes and Ive tested up to 20 minutes of audio.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Minimum <span style="color: red;">18GB</span> free disk space (most of it is used temporarily).
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 There is no major need to chop down your audio files into small slices as Step 1 will do that for you automatically and prepare the training set. Ive been testing with 5 minute long clips.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Some decent quality audio, multiple files if you like. Minimum of 2 minutes and Ive tested up to 20 minutes of audio.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 This process will need access to all your GPU and VRAM, so close any other software that's using your GPU currently.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - There is no major need to chop down your audio files into small slices as Step 1 will do that for you automatically and prepare the training set. Ive been testing with 5 minute long clips.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - This process will need access to all your GPU and VRAM, so close any other software that's using your GPU currently.
                 ### 🟨 <u>Setting up cuBLAS and cuDNN <span style="color: red;">Version 11.8</span></u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 If you have the <span style="color: #3366ff;;">Nvidia CUDA Toolkit Version 11.8</span> installed and can type <span style="color: #3366ff;;">nvcc --version</span> at the command prompt/terminal and it reports <span style="color: #00a000;">Cuda compilation tools, release 11.8</span> you should be good to go. 
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - If you have the <span style="color: #3366ff;;">Nvidia CUDA Toolkit Version 11.8</span> installed and can type <span style="color: #3366ff;;">nvcc --version</span> at the command prompt/terminal and it reports <span style="color: #00a000;">Cuda compilation tools, release 11.8</span> you should be good to go. 
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 If you dont have the toolkit installed, the idea is just to install the smallest bit possible and this will not affect or impact other things on your system.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - If you dont have the toolkit installed, the idea is just to install the smallest bit possible and this will not affect or impact other things on your system.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 You will need to download the Nvidia Cuda Toolkit 11.8<span style="color: #3366ff;"> network</span> install from [here](https://developer.nvidia.com/cuda-11-8-0-download-archive) 
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - You will need to download the Nvidia Cuda Toolkit 11.8<span style="color: #3366ff;"> network</span> install from [here](https://developer.nvidia.com/cuda-11-8-0-download-archive) 
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 1) Run the installer and select <span style="color: #3366ff;">Custom Advanced</span> Uncheck <span style="color: #3366ff;">everything</span> at the top then expand <span style="color: #3366ff;">CUDA</span>, <span style="color: #3366ff;">Development</span> > <span style="color: #3366ff;">Compiler</span> > and select <span style="color: #3366ff;;">nvcc</span> then expand <span style="color: #3366ff;;">Libraries</span> and select <span style="color: #3366ff;;">CUBLAS</span>.
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 2) Back at the top of <span style="color: #3366ff;">CUDA</span>, expand <span style="color: #3366ff;">Runtime</span> > <span style="color: #3366ff;">Libraries</span> and select <span style="color: #3366ff;">CUBLAS</span>. Click <span style="color: #3366ff;;">Next</span>, accept the default path (taking a note of its location) and let the install run. 
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 3) You should be able to drop to your terminal or command prompt and type <span style="color: #3366ff;">nvcc --version</span> and have it report <span style="color: #00a000;">Cuda compilation tools, release 11.8</span>. If it does you are good to go. If it doesn't > Step 4.
@@ -766,16 +913,16 @@ if __name__ == "__main__":
                 f"""
                 ## <u>STEP 1 - Preparing Audio/Generating the dataset</u><br>
                 ### 🟦 <u>What you need to do</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Please read Coqui's guide on what makes a good dataset [here](https://docs.coqui.ai/en/latest/what_makes_a_good_dataset.html#what-makes-a-good-dataset)
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Please read Coqui's guide on what makes a good dataset [here](https://docs.coqui.ai/en/latest/what_makes_a_good_dataset.html#what-makes-a-good-dataset)
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Place your audio files in <span style="color: #3366ff;">{str(audio_folder)}</span>                
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Place your audio files in <span style="color: #3366ff;">{str(audio_folder)}</span>                
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Your audio samples can be in the format <span style="color: #3366ff;">mp3, wav,</span> or <span style="color: #3366ff;">flac.</span>
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Your audio samples can be in the format <span style="color: #3366ff;">mp3, wav,</span> or <span style="color: #3366ff;">flac.</span>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 You will need a minimum of <span style="color: #3366ff;">2 minutes</span> of audio in either one or multiple audio files. 
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - You will need a minimum of <span style="color: #3366ff;">2 minutes</span> of audio in either one or multiple audio files. Very small sample files cause errors, so I would suggest 30 seconds and longer samples. 
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 When you have completed Steps 1, 2, and 3, you are welcome to delete your samples from "put-voice-samples-in-here".<br>
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - When you have completed Steps 1, 2, and 3, you are welcome to delete your samples from "put-voice-samples-in-here".<br>
                 ### 🟨 <u>What this step is doing</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 With step one, we are going to be stripping your audio file(s) into smaller files, using Whisper to find spoken words/sentences, compile that into excel sheets of training data, ready for finetuning the model on Step 2.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - With step one, we are going to be stripping your audio file(s) into smaller files, using Whisper to find spoken words/sentences, compile that into excel sheets of training data, ready for finetuning the model on Step 2.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 Whilst you can choose multiple Whisper models, its best only to use the 1 model as each one is about 3GB in size and will download to your local huggingface cache on first-time use. If and when you have completed training, you wish to delete this 3GB model from your system, you are welcome to do so.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Whilst you can choose multiple Whisper models, its best only to use the 1 model as each one is about 3GB in size and will download to your local huggingface cache on first-time use. If and when you have completed training, you wish to delete this 3GB model from your system, you are welcome to do so.
                 ### 🟩 <u>How long will this take?</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟩 First time, it needs to download the Whisper model which is 3GB. After that a few minutes on an average 3-4 year old system.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - First time, it needs to download the Whisper model which is 3GB. After that a few minutes on an average 3-4 year old system.
                 """
             )
                  
@@ -784,41 +931,41 @@ if __name__ == "__main__":
                 value=out_path,
                 visible=False,
             )
+            with gr.Row():
+                whisper_model = gr.Dropdown(
+                    label="Whisper Model",
+                    value="large-v3",
+                    choices=[
+                        "large-v3",
+                        "large-v2",
+                        "large",
+                        "medium",
+                        "small"
+                    ],
+                )
 
-            whisper_model = gr.Dropdown(
+                lang = gr.Dropdown(
-                label="Whisper Model",
+                    label="Dataset Language",
-                value="large-v3",
+                    value="en",
-                choices=[
+                    choices=[
-                    "large-v3",
+                        "en",
-                    "large-v2",
+                        "es",
-                    "large",
+                        "fr",
-                    "medium",
+                        "de",
-                    "small"
+                        "it",
-                ],
+                        "pt",
-            )
+                        "pl",
-
+                        "tr",
-            lang = gr.Dropdown(
+                        "ru",
-                label="Dataset Language",
+                        "nl",
-                value="en",
+                        "cs",
-                choices=[
+                        "ar",
-                    "en",
+                        "zh",
-                    "es",
+                        "hu",
-                    "fr",
+                        "ko",
-                    "de",
+                        "ja"
-                    "it",
+                    ],
-                    "pt",
+                )
-                    "pl",
-                    "tr",
-                    "ru",
-                    "nl",
-                    "cs",
-                    "ar",
-                    "zh",
-                    "hu",
-                    "ko",
-                    "ja"
-                ],
-            )
             progress_data = gr.Label(
                 label="Progress:"
             )
@@ -863,51 +1010,67 @@ if __name__ == "__main__":
                 f"""
                 ## <u>STEP 2 - Fine-tuning the XTTS Encoder</u><br>
                 ### 🟦 <u>What you need to do</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 The <span style="color: #3366ff;">Train CSV</span> and <span style="color: #3366ff;">Eval CSV</span> should already be populated. If not, just go back to Step 1 and click the button again.             
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - The <span style="color: #3366ff;">Train CSV</span> and <span style="color: #3366ff;">Eval CSV</span> should already be populated. If not, just go back to Step 1 and click the button again.             
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 If for any reason it crashed, close the app and re-start the finetuning and kick it off again, clicking through setp 1 (it doesnt need to regenerate the data)
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - If for any reason it crashed, close the app and re-start the finetuning and kick it off again, clicking through setp 1 (it doesnt need to regenerate the data)
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 The settings below are factory and should be ok.<br>
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - The settings below are factory and should be ok.<br>
                 ### 🟨 <u>What this step is doing</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 Very simply put, it's taking all our wav files generated in Step 1, along with our recorded speech in out excel documents and its training the model on that voice.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Very simply put, it's taking all our wav files generated in Step 1, along with our recorded speech in out excel documents and its training the model on that voice.
                 ### 🟩 <u>How long will this take?</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟩 On a RTX 4070 with factory settings (as below) and 20 minutes of audio, this took 20 minutes. Again, it will vary by system and how much audio you are throwing at it. You have time to go make a coffee.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - On a RTX 4070 with factory settings (as below) and 20 minutes of audio, this took 20 minutes. Again, it will vary by system and how much audio you are throwing at it. You have time to go make a coffee.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟩 Look at your command prompt/terminal window if you want to see what it is doing.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Look at your command prompt/terminal window if you want to see what it is doing.
                 """
             )
+            with gr.Row():
+                train_csv = gr.Textbox(
+                    label="Train CSV:",
+                )
+                eval_csv = gr.Textbox(
+                    label="Eval CSV:",
+                )
+            with gr.Row():
+                num_epochs =  gr.Slider(
+                    label="Number of epochs:",
+                    minimum=1,
+                    maximum=100,
+                    step=1,
+                    value=args.num_epochs,
+                )
+                batch_size = gr.Slider(
+                    label="Batch size:",
+                    minimum=2,
+                    maximum=512,
+                    step=1,
+                    value=args.batch_size,
+                )
+                grad_acumm = gr.Slider(
+                    label="Grad accumulation steps:",
+                    minimum=2,
+                    maximum=128,
+                    step=1,
+                    value=args.grad_acumm,
+                )
+                max_audio_length = gr.Slider(
+                    label="Max permitted audio size in seconds:",
+                    minimum=2,
+                    maximum=20,
+                    step=1,
+                    value=args.max_audio_length,
+                )
+            with gr.Row():
+                model_to_train_choices = ["Base Model"]
+                if finetuned_model:
+                    model_to_train_choices.append("Existing finetuned model")
+                model_to_train = gr.Radio(
+                    choices=model_to_train_choices,
+                    label="Which model do you want to train?",
+                    value="Base Model"
+                )
+                gr.Markdown(
+                f"""
+                ##### If you have an existing finetuned model in <span style="color: #3366ff;">/models/trainedmodel/</span> you will have an option to select <span style="color: #3366ff;">Existing finetuned model</span>, allowing you to further train it. Otherwise, you will only have Base Model as an option.
+                """
+                )
 
-            train_csv = gr.Textbox(
-                label="Train CSV:",
-            )
-            eval_csv = gr.Textbox(
-                label="Eval CSV:",
-            )
-            num_epochs =  gr.Slider(
-                label="Number of epochs:",
-                minimum=1,
-                maximum=100,
-                step=1,
-                value=args.num_epochs,
-            )
-            batch_size = gr.Slider(
-                label="Batch size:",
-                minimum=2,
-                maximum=512,
-                step=1,
-                value=args.batch_size,
-            )
-            grad_acumm = gr.Slider(
-                label="Grad accumulation steps:",
-                minimum=2,
-                maximum=128,
-                step=1,
-                value=args.grad_acumm,
-            )
-            max_audio_length = gr.Slider(
-                label="Max permitted audio size in seconds:",
-                minimum=2,
-                maximum=20,
-                step=1,
-                value=args.max_audio_length,
-            )
             progress_train = gr.Label(
                 label="Progress:"
             )
@@ -949,16 +1112,16 @@ if __name__ == "__main__":
                 f"""
                 ## <u>STEP 3 -  Inference/Testing</u><br>
                 ### 🟥 <u>Important Note</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟥 This step will error if you are using TTS version 0.22.0. Please re-run  <span style="color: #3366ff;">pip install -r requirements_finetune.txt</span> if it errors loading the model.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - This step will error if you are using TTS version 0.22.0. Please re-run  <span style="color: #3366ff;">pip install -r requirements_finetune.txt</span> if it errors loading the model.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟥 If you dont see multiple speaker reference files and you used more than 3 minutes of speech, try refreshing the page as it may not have loaded them correctly.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - If you dont see multiple speaker reference files and you used more than 3 minutes of speech, try refreshing the page as it may not have loaded them correctly.
                 ### 🟦 <u>What you need to do</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 The model is now trained and you are at the testing stage. Hopefully all the dropdowns should be pre-populated now.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - The model is now trained and you are at the testing stage. Hopefully all the dropdowns should be pre-populated now.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 You need to <span style="color: #3366ff;">Load Fine-tuned XTTS model</span> and then select your <span style="color: #3366ff;">Speaker Reference Audio</span>. You can choose various <span style="color: #3366ff;">Speaker Reference Audios</span> to see which works best.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - You need to <span style="color: #3366ff;">Load Fine-tuned XTTS model</span> and then select your <span style="color: #3366ff;">Speaker Reference Audio</span>. You can choose various <span style="color: #3366ff;">Speaker Reference Audios</span> to see which works best.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 All the <span style="color: #3366ff;">Speaker Reference Audios</span> in the dropdown are ones that are <span style="color: #3366ff;">8 seconds</span> long or more. You can use one of these later for your voice sample in All Talk, so remember the one you like.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - All the <span style="color: #3366ff;">Speaker Reference Audios</span> in the dropdown are ones that are <span style="color: #3366ff;">8 seconds</span> long or more. You can use one of these later for your voice sample in All Talk, so remember the one you like.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Some <span style="color: #3366ff;">Speaker Reference Audios</span> are a bit long and may cause issues generating the TTS, so try different ones.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Some <span style="color: #3366ff;">Speaker Reference Audios</span> are a bit long and may cause issues generating the TTS, so try different ones.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 Once you are happy, move to the next tab for instruction on cleaning out the training data and how to copy your newly trained model over your current model.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Once you are happy, move to the next tab for instruction on cleaning out the training data and how to copy your newly trained model over your current model.
                 ### 🟨 <u>What this step is doing</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 Its loading the finetuned model into memory, loading a voice sample and generating TTS, so that you can test out how well fine tuning worked.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Its loading the finetuned model into memory, loading a voice sample and generating TTS, so that you can test out how well fine tuning worked.
                 """
             )
 
@@ -1051,25 +1214,57 @@ if __name__ == "__main__":
                 f"""
                 ## <u>What to do next</u><br>
                 ### 🟦 <u>What you need to do</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 There are 3x buttons below:
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - You have a few options below:
-                #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 🔹 Compact &amp; move model. This will compress the trained model and move it, along with any large enough wav files created to <span style="color: #3366ff;">/models/trainedmodel/</span>. You can use the wav files as sample voices..
+                #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 🔹 Compact &amp; move model. This will compress the raw finetuned model and move it, along with any large enough wav files created to one of two locations (please read text next to the button).<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; You can use the wav files in that location as sample voices with your model. Models moved to <span style="color: #3366ff;">/models/trainedmodel/</span> will be loadable within Text-gen-webui on the <span style="color: #3366ff;">XTTSv2 FT</span> loader.
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 🔹 Delete Generated Training data. This will delete the finetuned model and any other training data generated during this process. You will more than likely want to do this after you have Compacted &amp; moved the model.
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 🔹 Delete original voice samples. This will delete the voice samples you placed in put-voice-samples-in-here (if you no longer have a use for them).
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 I will at some point add an option within AllTalk to load a finetuned model as a seperate model. However, for now, you will have to copy it over the model in the <span style="color: #3366ff;">/models/xttsv2_2.0.2/</span> model, if you wish to use it.
+                #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; 🔹 Compress a legacy finetuned model. For any models that were generated pre the current version of finetuning, you can still compact them down.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟦 If you wish to further train this model, you will also need to copy it over the model in the <span style="color: #3366ff;">/models/xttsv2_2.0.2/</span> model folder.. 
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - Instructions on using the finetuned model in Text-generation-webui and also further tuning this model you just created are on <span style="color: #3366ff;">OPTION A - Compact and move model to '/trainedmodel/'</span>.
                 ### 🟨 <u>Clearing up disk space</u>
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 If you are not going to train anything again, you can delete the whisper model from inside of your huggingface cache (3GB approx) 
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - If you are not going to train anything again, you can delete the whisper model from inside of your huggingface cache (3GB approx) 
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Linux <span style="color: #3366ff;">~/.cache/huggingface/hub/(folder-here)</span>
                 #### &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Windows <span style="color: #3366ff;">C:&bsol;users&lpar;your-username&rpar;&bsol;.cache&bsol;huggingface&bsol;hub&bsol;(folder-here)</span>.
-                #### &nbsp;&nbsp;&nbsp;&nbsp; 🟨 You can also uninstall the Nvidia CUDA 11.8 Toolkit if you wish and remove your environment variable entries.
+                #### &nbsp;&nbsp;&nbsp;&nbsp; - You can also uninstall the Nvidia CUDA 11.8 Toolkit if you wish and remove your environment variable entries.
                 """
             )
             final_progress_data = gr.Label(
                 label="Progress:"
             )
-            compact_btn = gr.Button(value="Compact and move model to '/models/trainedmodel/' (THIS WILL OVER-WRITE ANY MODEL IN THAT LOCATION)")
+            with gr.Row():
-            delete_training_btn = gr.Button(value="Delete generated training data")
+                compact_btn = gr.Button(value="OPTION A - Compact and move model to '/trainedmodel/'")
-            delete_voicesamples_btn = gr.Button(value="Delete original voice samples")
+                gr.Markdown(
+                f"""
+                ##### This will compact the model and move it, along with the reference wav's to <span style="color: #3366ff;">/models/trainedmodel/</span> and will <span style="color: red;">OVERWRITE</span> any model in that location. This model will become available to load in Text-gen-webui with the <span style="color: #3366ff;">XTTSv2 FT</span> loader. It will also be detected by Finetune as a model that can be further trained (Reload finetune.py).
+                """
+                )
+            with gr.Row():
+                compact_lastfinetuned_btn = gr.Button(value="OPTION B - Compact and move model to '/lastfinetuned/'")
+                gr.Markdown(
+                f"""
+                ##### This will compact the model and move it, along with the reference wav's to <span style="color: #3366ff;">/models/lastfinetuned/</span> and will <span style="color: red;">OVERWRITE</span> any model in that location. This will just leave you a finetuned model that you can do with as you please. It will <span style="color: red;">NOT</span> be available in Text-gen-webui unless you copy it over one of the other model loader locations.
+                """
+                )
+            with gr.Row():
+                delete_training_btn = gr.Button(value="Delete generated training data")
+                gr.Markdown(
+                f"""
+                ##### This will <span style="color: red;">DELETE</span> your training data and the raw finetuned model from <span style="color: #3366ff;">/finetune/tmp-trn/</span>.
+                """
+                )
+            with gr.Row():
+                delete_voicesamples_btn = gr.Button(value="Delete original voice samples")
+                gr.Markdown(
+                f"""
+                ##### This will <span style="color: red;">DELETE</span> your original voice samples from <span style="color: #3366ff;">/finetune/put-voice-samples-in-here/</span>.
+                """
+                )
+            with gr.Row():
+                compact_legacy_btn = gr.Button(value="Compact a legacy finetuned model")
+                gr.Markdown(
+                f"""
+                ##### This will compact your old finetuned models from 5GB to 1.9GB. Place your 5GB <span style="color: #3366ff;">base_model.pth</span> file in <span style="color: #3366ff;">/finetune/</span> (rename your file to <span style="color: #3366ff;">base_model.pth</span> if necessary). This will generate you a <span style="color: #3366ff;">model.pth</span> file.
+                """
+                )
 
             prompt_compute_btn.click(
                 fn=preprocess_dataset,
@@ -1125,6 +1320,10 @@ if __name__ == "__main__":
                 fn=compact_model,
                 outputs=[final_progress_data],
             )
+            compact_lastfinetuned_btn.click(
+                fn=compact_lastfinetuned_model,
+                outputs=[final_progress_data],
+            )
             delete_training_btn.click(
                 fn=delete_training_data,
                 outputs=[final_progress_data],
@@ -1133,6 +1332,11 @@ if __name__ == "__main__":
                 fn=delete_voice_sample_contents,
                 outputs=[final_progress_data],
             )
+            compact_legacy_btn.click(
+                fn=compact_legacy_model,
+                outputs=[final_progress_data],
+            )
+            model_to_train.change(basemodel_or_finetunedmodel_choice, model_to_train, None)
 
     demo.queue().launch(
         show_api=False,

modeldownload.json

@@ -8,6 +8,7 @@
         "model.pth": "https://huggingface.co/coqui/XTTS-v2/resolve/v2.0.2/model.pth?download=true",
         "dvae.pth": "https://huggingface.co/coqui/XTTS-v2/resolve/v2.0.2/dvae.pth?download=true",
         "mel_stats.pth": "https://huggingface.co/coqui/XTTS-v2/resolve/v2.0.2/mel_stats.pth?download=true",
+        "speakers_xtts.pth": "https://huggingface.co/coqui/XTTS-v2/resolve/v2.0.2/speakers_xtts.pth?download=true",
         "vocab.json": "https://huggingface.co/coqui/XTTS-v2/resolve/v2.0.2/vocab.json?download=true"
     }
 }

requirements_nvidia.txt

@@ -11,3 +11,5 @@ tqdm>=4.66.1
 importlib-metadata>=4.8.1
 packaging>=23.2
 pydantic>=1.10.13
+sounddevice>=0.4.6
+python-multipart>=0.0.6

requirements_other.txt

@@ -11,3 +11,5 @@ tqdm>=4.66.1
 importlib-metadata>=4.8.1
 packaging>=23.2
 pydantic>=1.10.13
+sounddevice>=0.4.6
+python-multipart>=0.0.6

script.py

@@ -77,6 +77,8 @@ with open(this_dir / "languages.json", encoding="utf8") as f:
     languages = json.load(f)
 # Create a global lock
 process_lock = threading.Lock()
+# Base setting for a possible FineTuned model existing and the loader being available
+tts_method_xtts_ft = False
 
 # Gather the voice files
 def get_available_voices():
@@ -139,10 +141,23 @@ def check_required_files():
     subprocess.run(["python", str(download_script_path)])
     print(f"[{params['branding']}Startup] All required files are present.")
 
-
 # STARTUP Call Check routine
 check_required_files()
 
+##################################################
+#### Check to see if a finetuned model exists ####
+##################################################
+# Set the path to the directory
+trained_model_directory = this_dir / "models" / "trainedmodel"
+# Check if the directory "trainedmodel" exists
+finetuned_model = trained_model_directory.exists()
+# If the directory exists, check for the existence of the required files
+# If true, this will add a extra option in the Gradio interface for loading Xttsv2 FT
+if finetuned_model:
+    required_files = ["model.pth", "config.json", "vocab.json"]
+    finetuned_model = all((trained_model_directory / file).exists() for file in required_files)
+if finetuned_model:
+    print(f"[{params['branding']}Startup] Finetuned model \033[93mDetected\033[0m")
 
 ####################################################
 #### SET GRADIO BUTTONS BASED ON confignew.json ####
@@ -252,7 +267,7 @@ else:
 timeout = 120  # Adjust the timeout as needed
 
 # Introduce a delay before starting the check loop
-time.sleep(25)  # Wait 25 secs before checking if the tts_server.py has started up.
+time.sleep(26)  # Wait 26 secs before checking if the tts_server.py has started up.
 start_time = time.time()
 while time.time() - start_time < timeout:
     try:
@@ -264,7 +279,7 @@ while time.time() - start_time < timeout:
         print(
             f"[{params['branding']}Startup] \033[91mWarning\033[0m TTS Subprocess has NOT started up yet, Will keep trying for 120 seconds maximum. Please wait."
         )
-    time.sleep(1)
+    time.sleep(4)
 else:
     print(
         f"[{params['branding']}Startup] Startup timed out. Check the server logs for more information."
@@ -272,11 +287,13 @@ else:
     # Cleanly kill off this script, but allow text-generation-webui to keep running, albeit without this alltalk_tts
     sys.exit(1)
 
+
 #####################################
 #### MODEL LOADING AND UNLOADING ####
 #####################################
 # MODEL - Swap model based on Gradio selection API TTS, API Local, XTTSv2 Local
 def send_reload_request(tts_method):
+    global tts_method_xtts_ft
     try:
         params["tts_model_loaded"] = False
         url = f"{base_url}/api/reload"
@@ -295,17 +312,26 @@ def send_reload_request(tts_method):
                 params["tts_method_api_tts"] = True
                 params["deepspeed_activate"] = False
                 audio_path = this_dir / "templates" / "apitts.wav"
+                tts_method_xtts_ft = False
             elif tts_method == "API Local":
                 params["tts_method_api_tts"] = False
                 params["tts_method_xtts_local"] = False
                 params["tts_method_api_local"] = True
                 params["deepspeed_activate"] = False
                 audio_path = this_dir / "templates" / "apilocal.wav"
+                tts_method_xtts_ft = False
             elif tts_method == "XTTSv2 Local":
                 params["tts_method_api_tts"] = False
                 params["tts_method_api_local"] = False
                 params["tts_method_xtts_local"] = True
                 audio_path = this_dir / "templates" / "xttslocal.wav"
+                tts_method_xtts_ft = False
+            elif tts_method == "XTTSv2 FT":
+                params["tts_method_api_tts"] = False
+                params["tts_method_api_local"] = False
+                params["tts_method_xtts_local"] = False
+                audio_path = this_dir / "templates" / "xttsfinetuned.wav"
+                tts_method_xtts_ft = True
         return f'<audio src="file/{audio_path}" controls autoplay></audio>'
     except requests.exceptions.RequestException as e:
         # Handle the HTTP request error
@@ -494,7 +520,6 @@ def output_modifier(string, state):
     if not params["activate"]:
         return string
     original_string = string
-    print("ORGINAL STRING: ", original_string)
     cleaned_string = before_audio_generation(string, params)
     if cleaned_string is None:
         return
@@ -505,7 +530,6 @@ def output_modifier(string, state):
     if process_lock.acquire(blocking=False):
         try:
             if params["narrator_enabled"]:
-                #print(original_string)
                 # Do Some basic stripping and remove any double CR's
                 processed_string = (
                     original_string
@@ -786,8 +810,11 @@ def ui():
             deepspeed_checkbox_play = gr.HTML(visible=False)
 
         with gr.Row():
+            model_loader_choices=["API TTS", "API Local", "XTTSv2 Local"]
+            if finetuned_model:
+                model_loader_choices.append("XTTSv2 FT")
             tts_radio_buttons = gr.Radio(
-                choices=["API TTS", "API Local", "XTTSv2 Local"],
+                choices=model_loader_choices,
                 label="Select TTS Generation Method (Read NOTE)",
                 value=gr_modelchoice,  # Set the default value
             )

tts_server.py

@@ -33,7 +33,8 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 # STARTUP VARIABLE - Import languges file for Gradio to be able to display them in the interface
 with open(this_dir / "languages.json", encoding="utf8") as f:
     languages = json.load(f)
-
+# Base setting for a possible FineTuned model existing and the loader being available
+tts_method_xtts_ft = False
 
 #################################################################
 #### LOAD PARAMS FROM confignew.json - REQUIRED FOR BRANDING ####
@@ -123,7 +124,7 @@ app = FastAPI(lifespan=startup_shutdown)
 #####################################
 #### MODEL LOADING AND UNLOADING ####
 #####################################
-# MODEL LOADERS Picker For API TTS, API Local, XTTSv2 Local
+# MODEL LOADERS Picker For API TTS, API Local, XTTSv2 Local, XTTSv2 FT
 async def setup():
     global device
     # Set a timer to calculate load times
@@ -150,6 +151,13 @@ async def setup():
             "\033[0m",
         )
         model = await xtts_manual_load_model()
+    elif tts_method_xtts_ft:
+        print(
+            f"[{params['branding']}Model] \033[94mXTTSv2 FT Loading\033[0m /models/fintuned/model.pth \033[94minto\033[93m",
+            device,
+            "\033[0m",
+        )
+        model = await xtts_ft_manual_load_model()
     # Create an end timer for calculating load times
     generate_end_time = time.time()
     # Calculate start time minus end time
@@ -222,6 +230,23 @@ async def xtts_manual_load_model():
     model.to(device)
     return model
 
+# MODEL LOADER For "XTTSv2 FT"
+async def xtts_ft_manual_load_model():
+    global model
+    config = XttsConfig()
+    config_path = this_dir / "models" / "trainedmodel" / "config.json"
+    vocab_path_dir = this_dir / "models" / "trainedmodel" / "vocab.json"
+    checkpoint_dir = this_dir / "models" / "trainedmodel"
+    config.load_json(str(config_path))
+    model = Xtts.init_from_config(config)
+    model.load_checkpoint(
+        config,
+        checkpoint_dir=str(checkpoint_dir),
+        vocab_path=str(vocab_path_dir),
+        use_deepspeed=params["deepspeed_activate"],
+    )
+    model.to(device)
+    return model
 
 # MODEL UNLOADER
 async def unload_model(model):
@@ -236,6 +261,7 @@ async def unload_model(model):
 # MODEL - Swap model based on Gradio selection API TTS, API Local, XTTSv2 Local
 async def handle_tts_method_change(tts_method):
     global model
+    global tts_method_xtts_ft
     # Update the params dictionary based on the selected radio button
     print(
         f"[{params['branding']}Model] \033[94mChanging model \033[92m(Please wait 15 seconds)\033[0m"
@@ -246,15 +272,23 @@ async def handle_tts_method_change(tts_method):
         params["tts_method_xtts_local"] = False
         params["tts_method_api_tts"] = True
         params["deepspeed_activate"] = False
+        tts_method_xtts_ft = False
     elif tts_method == "API Local":
         params["tts_method_api_tts"] = False
         params["tts_method_xtts_local"] = False
         params["tts_method_api_local"] = True
         params["deepspeed_activate"] = False
+        tts_method_xtts_ft = False
     elif tts_method == "XTTSv2 Local":
         params["tts_method_api_tts"] = False
         params["tts_method_api_local"] = False
         params["tts_method_xtts_local"] = True
+        tts_method_xtts_ft = False
+    elif tts_method == "XTTSv2 FT":
+        tts_method_xtts_ft = True
+        params["tts_method_api_tts"] = False
+        params["tts_method_api_local"] = False
+        params["tts_method_xtts_local"] = False
 
     # Unload the current model
     model = await unload_model(model)
@@ -267,7 +301,7 @@ async def handle_tts_method_change(tts_method):
 @app.route("/api/reload", methods=["POST"])
 async def reload(request: Request):
     tts_method = request.query_params.get("tts_method")
-    if tts_method not in ["API TTS", "API Local", "XTTSv2 Local"]:
+    if tts_method not in ["API TTS", "API Local", "XTTSv2 Local", "XTTSv2 FT"]:
         return {"status": "error", "message": "Invalid TTS method specified"}
     await handle_tts_method_change(tts_method)
     return Response(
@@ -424,8 +458,8 @@ async def generate_audio(text, voice, language, output_file):
     if params["low_vram"] and device == "cpu":
         await switch_device()
     generate_start_time = time.time()  # Record the start time of generating TTS
-    # XTTSv2 LOCAL Method
+    # XTTSv2 LOCAL & Xttsv2 FT Method
-    if params["tts_method_xtts_local"]:
+    if params["tts_method_xtts_local"] or tts_method_xtts_ft:
         print(f"[{params['branding']}TTSGen] {text}")
         gpt_cond_latent, speaker_embedding = model.get_conditioning_latents(
             audio_path=[f"{this_dir}/voices/{voice}"],
@@ -619,6 +653,53 @@ async def get_audio(filename: str):
     audio_path = this_dir / "outputs" / filename
     return FileResponse(audio_path)
 
+
+#########################
+#### VOICES LIST API ####
+#########################
+# Define the new endpoint
+@app.get("/api/voices")
+async def get_voices():
+    wav_files = list_files(this_dir / "voices")
+    return {"voices": wav_files}
+
+###########################
+#### PREVIEW VOICE API ####
+###########################
+@app.post("/api/previewvoice/", response_class=JSONResponse)
+async def preview_voice(request: Request, voice: str = Form(...)):
+    try:
+        # Hardcoded settings
+        language = "en"
+        output_file_name = "api_preview_voice"
+
+        # Clean the voice filename for inclusion in the text
+        clean_voice_filename = re.sub(r'\.wav$', '', voice.replace(' ', '_'))
+        clean_voice_filename = re.sub(r'[^a-zA-Z0-9]', ' ', clean_voice_filename)
+        
+        # Generate the audio
+        text = f"Hello, this is a preview of voice {clean_voice_filename}."
+
+        # Generate the audio
+        output_file_path = this_dir / "outputs" / f"{output_file_name}.wav"
+        await generate_audio(text, voice, language, output_file_path)
+
+        # Generate the URL
+        output_file_url = f'http://{params["ip_address"]}:{params["port_number"]}/audio/{output_file_name}.wav'
+
+        # Return the response with both local file path and URL
+        return JSONResponse(
+            content={
+                "status": "generate-success",
+                "output_file_path": str(output_file_path),
+                "output_file_url": str(output_file_url),
+            },
+            status_code=200,
+        )
+    except Exception as e:
+        print(f"An error occurred: {e}")
+        return JSONResponse(content={"error": "An error occurred"}, status_code=500)
+
 ########################
 #### GENERATION API ####
 ########################
@@ -627,9 +708,29 @@ import re
 import uuid
 import numpy as np
 import soundfile as sf
+import sys
+
+# Check for PortAudio library on Linux
+try:
+    import sounddevice as sd
+    sounddevice_installed=True
+except OSError:
+    print(f"[{params['branding']}Startup] \033[91mInfo\033[0m PortAudio library not found. If you wish to play TTS in standalone mode through the API suite")
+    print(f"[{params['branding']}Startup] \033[91mInfo\033[0m please install PortAudio. This will not affect any other features or use of Alltalk.")
+    print(f"[{params['branding']}Startup] \033[91mInfo\033[0m If you don't know what the API suite is, then this message is nothing to worry about.")
+    sounddevice_installed=False
+    if sys.platform.startswith('linux'):
+        print(f"[{params['branding']}Startup] \033[91mInfo\033[0m On Linux, you can use the following command to install PortAudio:")
+        print(f"[{params['branding']}Startup] \033[91mInfo\033[0m sudo apt-get install portaudio19-dev")
+
 from typing import Union, Dict
 from pydantic import BaseModel, ValidationError, Field
 
+def play_audio(file_path, volume):
+    data, fs = sf.read(file_path)
+    sd.play(volume * data, fs)
+    sd.wait()
+
 class Request(BaseModel):
     # Define the structure of the 'Request' class if needed
     pass
@@ -738,10 +839,6 @@ async def tts_generate(
     autoplay_volume: float = Form(...),
 ):
     try:
-        #print(f"text_filtering: {text_filtering}")
-        #print(f"narrator_enabled: {narrator_enabled}")
-        #print(f"text_not_inside: {text_not_inside}")
-        #print(f"output_file_timestamp: {output_file_timestamp}")
         json_input_data = {
             "text_input": text_input,
             "text_filtering": text_filtering,
@@ -761,46 +858,38 @@ async def tts_generate(
         else:
             return JSONResponse(content={"error": JSONresult}, status_code=400)
         if narrator_enabled:
-            print("ORIGINAL UNTOUCHED STRING IS:", text_input,"\n")
+            #print("ORIGINAL UNTOUCHED STRING IS:", text_input,"\n")
             if text_filtering in ["standard", "none"]:
                     cleaned_string = (
                         text_input
                         .replace('"', '"<')
                         .replace('*', '<*')
                     )
-                    print("STANDARD FILTERING IS NOW:", cleaned_string,"\n")
-                    #parts = re.split(r'\.(?<!<[*"])', cleaned_string)
                     parts = re.split(r'(?<=<\*\s)|(?<=\.\s)|(?<=\."\<)|(?<=\."<)', cleaned_string)
                     parts = list(filter(lambda x: x.strip(), parts))
             elif text_filtering == "html":
                 cleaned_string = standard_filtering(text_input)
                 cleaned_string = narrator_filtering(cleaned_string)
-                print("HTML FILTERING IS NOW:", cleaned_string,"\n")
                 parts = re.split(r'&quot;|\.\*', cleaned_string)
             audio_files_all_paragraphs = []
             audio_files_paragraph = []
             for i, part in enumerate(parts):
                 if len(part.strip()) <= 1:
                     continue
-                print("THIS IS A PART", part)
                 # Figure out which type of line it is, then replace characters as necessary to avoid TTS trying to pronunce them, htmlunescape after. 
                 # Character will always be a < with a letter immediately after it
                 if '<' in part and '<*' not in part and '< *' not in part and '<  *' not in part and '< ' not in part and '<  ' not in part:
-                    print("IF - Character\n")
                     cleaned_part = html.unescape(part.replace('<', ''))
                     voice_to_use = character_voice_gen
                 #Narrator will always be an * or < with an * a position or two after it.
                 elif '<*' in part or '< *' in part or '<  *' in part or '*' in part:
-                    print("IF - Narrator\n")
                     cleaned_part = html.unescape(part.replace('<*', '').replace('< *', '').replace('<  *', '').replace('*', '').replace('<. ', '')) 
                     voice_to_use = narrator_voice_gen
                 #If the other two dont capture it, aka, the AI gave no * or &quot; on the line, use non_quoted_text_is aka user interface, user can choose Char or Narrator
                 elif text_not_inside == "character":
-                    print("ELSE - CHARACTER\n")
                     cleaned_part = html.unescape(part.replace('< ', '').replace('<  ', '').replace('<  ', ''))
                     voice_to_use = character_voice_gen
                 elif text_not_inside == "narrator":
-                    print("ELSE - NARRATOR\n")
                     cleaned_part = html.unescape(part.replace('< ', '').replace('<  ', '').replace('<  ', ''))
                     voice_to_use = narrator_voice_gen
                 output_file = this_dir / "outputs" / f"{output_file_name}_{uuid.uuid4()}_{int(time.time())}_{i}.wav"
@@ -827,6 +916,10 @@ async def tts_generate(
             else:
                 cleaned_string = text_input
             await generate_audio(cleaned_string, character_voice_gen, language, output_file_path)
+        if sounddevice_installed == False:
+            autoplay = False
+        if autoplay:
+            play_audio(output_file_path, autoplay_volume)
         return JSONResponse(content={"status": "generate-success", "output_file_path": str(output_file_path), "output_file_url": str(output_file_url)}, status_code=200)
     except Exception as e:
         return JSONResponse(content={"status": "generate-failure", "error": "An error occurred"}, status_code=500)
@@ -1673,6 +1766,33 @@ simple_webpage = """
 <p style="padding-left: 30px;"><span style="color: #3366ff;">curl -X POST "http://127.0.0.1:7851/api/lowvramsetting?new_low_vram_value=True"</span></p>
 <p style="padding-left: 30px;">Replace True with False to disable Low VRAM mode.</p>
 <p style="padding-left: 30px;">JSON return <span style="color: #339966;">{"status": "lowvram-success"}</span></p>
+<h4>Ready Endpoint</strong></h4>
+<p>Check if the Text-to-Speech (TTS) service is ready to accept requests.</p>
+<ul>
+  <li>URL: <span style="color: #3366ff;">http://127.0.0.1:7851/api/ready</span></li>
+  <li>Method: <span style="color: #3366ff;">GET</span></li>
+  <li>Response: <span style="color: #339966;">Ready</span></li>
+</ul>
+<p style="padding-left: 30px;"><span style="color: #3366ff;">curl -X GET "http://127.0.0.1:7851/api/ready"</span></p>
+<h4>Voices List Endpoint</strong></h4>
+<p>Retrieve a list of available voices for generating speech.</p>
+<ul>
+  <li>URL: <span style="color: #3366ff;">http://127.0.0.1:7851/api/voices</span></li>
+  <li>Method: <span style="color: #3366ff;">GET</span></li>
+</ul>
+<p style="padding-left: 30px;"><span style="color: #3366ff;">curl -X GET "http://127.0.0.1:7851/api/voices"</span></p>
+<p style="padding-left: 30px;">JSON return: <span style="color: #339966;">{"voices": ["voice1.wav", "voice2.wav", "voice3.wav"]}</span></p>
+<h4><strong>Preview Voice Endpoint</strong></h4>
+<p>Generate a preview of a specified voice with hardcoded settings.</p>
+<ul>
+  <li>URL: <span style="color: #3366ff;">http://127.0.0.1:7851/api/previewvoice/</span></li>
+  <li>Method: <span style="color: #3366ff;">POST</span></li>
+  <li>Content-Type: <span style="color: #3366ff;">application/x-www-form-urlencoded</span></li>
+</ul>
+<p style="padding-left: 30px;"><span style="color: #3366ff;">curl -X POST "http://127.0.0.1:7851/api/previewvoice/" -F "voice=female_01.wav"</span></p>
+<p style="padding-left: 30px;">Replace <span style="color: #3366ff;">female_01.wav</span> with the name of the voice sample you want to hear.</p>
+<p style="padding-left: 30px;">JSON return: <span style="color: #339966;">{"status": "generate-success", "output_file_path": "/path/to/outputs/api_preview_voice.wav", "output_file_url": "http://127.0.0.1:7851/audio/api_preview_voice.wav"}</span></p>
+
 <p><a href="#toc">Back to top of page<br /></a></p>
 
 <h2 id="debugging-and-tts-generation-information"><strong>Debugging and TTS Generation Information</strong></h2>
@@ -1717,10 +1837,20 @@ template = Template(simple_webpage)
 # Render the template with the dynamic values
 rendered_html = template.render(params=params)
 
+###############################
+#### Internal script ready ####
+###############################
 @app.get("/ready")
 async def ready():
     return Response("Ready endpoint")
 
+############################
+#### External API ready ####
+############################
+@app.get("/api/ready")
+async def ready():
+    return Response("Ready")
+
 @app.get("/")
 async def read_root():
     return HTMLResponse(content=rendered_html, status_code=200)