Sync from GitHub repo

This Space is synced from the GitHub repo: https://github.com/SWivid/F5-TTS. Please submit contributions to the Space there

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -0,0 +1,50 @@
+name: "Bug Report"
+description: |
+  Please provide as much details to help address the issue, including logs and screenshots.
+labels:
+  - bug
+body:
+  - type: checkboxes
+    attributes:
+      label: Checks
+      description: "To ensure timely help, please confirm the following:"
+      options:
+        - label: This template is only for bug reports, usage problems go with 'Help Wanted'.
+          required: true
+        - label: I have thoroughly reviewed the project documentation but couldn't find information to solve my problem.
+          required: true
+        - label: I have searched for existing issues, including closed ones, and couldn't find a solution.
+          required: true
+        - label: I confirm that I am using English to submit this report in order to facilitate communication.
+          required: true
+  - type: textarea
+    attributes:
+      label: Environment Details
+      description: "Provide details such as OS, Python version, and any relevant software or dependencies."
+      placeholder: e.g., CentOS Linux 7, RTX 3090, Python 3.10, torch==2.3.0, cuda 11.8
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: Steps to Reproduce
+      description: |
+        Include detailed steps, screenshots, and logs. Use the correct markdown syntax for code blocks.
+      placeholder: |
+        1. Create a new conda environment.
+        2. Clone the repository, install as local editable and properly set up.
+        3. Run the command: `accelerate launch src/f5_tts/train/train.py`.
+        4. Have following error message... (attach logs).
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: ✔️ Expected Behavior
+      placeholder: Describe what you expected to happen.
+    validations:
+      required: false
+  - type: textarea
+    attributes:
+      label: ❌ Actual Behavior
+      placeholder: Describe what actually happened.
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -0,0 +1,62 @@
+name: "Feature Request"
+description: |
+  Some constructive suggestions and new ideas regarding current repo.
+labels:
+  - enhancement
+body:
+  - type: checkboxes
+    attributes:
+      label: Checks
+      description: "To help us grasp quickly, please confirm the following:"
+      options:
+        - label: This template is only for feature request.
+          required: true
+        - label: I have thoroughly reviewed the project documentation but couldn't find any relevant information that meets my needs.
+          required: true
+        - label: I have searched for existing issues, including closed ones, and found not discussion yet.
+          required: true
+        - label: I confirm that I am using English to submit this report in order to facilitate communication.
+          required: true
+  - type: textarea
+    attributes:
+      label: 1. Is this request related to a challenge you're experiencing? Tell us your story.
+      description: |
+        Describe the specific problem or scenario you're facing in detail. For example:
+        *"I was trying to use [feature] for [specific task], but encountered [issue]. This was frustrating because...."*
+      placeholder: Please describe the situation in as much detail as possible.
+    validations:
+      required: true
+
+  - type: textarea
+    attributes:
+      label: 2. What is your suggested solution?
+      description: |
+        Provide a clear description of the feature or enhancement you'd like to propose. 
+        How would this feature solve your issue or improve the project?
+      placeholder: Describe your idea or proposed solution here.
+    validations:
+      required: true
+
+  - type: textarea
+    attributes:
+      label: 3. Additional context or comments
+      description: |
+        Any other relevant information, links, documents, or screenshots that provide clarity. 
+        Use this section for anything not covered above.
+      placeholder: Add any extra details here.
+    validations:
+      required: false
+
+  - type: checkboxes
+    attributes:
+      label: 4. Can you help us with this feature?
+      description: |
+        Let us know if you're interested in contributing. This is not a commitment but a way to express interest in collaboration.
+      options:
+        - label: I am interested in contributing to this feature.
+          required: false
+
+  - type: markdown
+    attributes:
+      value: |
+        **Note:** Please submit only one request per issue to keep discussions focused and manageable.

.github/ISSUE_TEMPLATE/help_wanted.yml

@@ -0,0 +1,50 @@
+name: "Help Wanted"
+description: |
+  Please provide as much details to help address the issue, including logs and screenshots.
+labels:
+  - help wanted
+body:
+  - type: checkboxes
+    attributes:
+      label: Checks
+      description: "To ensure timely help, please confirm the following:"
+      options:
+        - label: This template is only for usage issues encountered.
+          required: true
+        - label: I have thoroughly reviewed the project documentation but couldn't find information to solve my problem.
+          required: true
+        - label: I have searched for existing issues, including closed ones, and couldn't find a solution.
+          required: true
+        - label: I confirm that I am using English to submit this report in order to facilitate communication.
+          required: true
+  - type: textarea
+    attributes:
+      label: Environment Details
+      description: "Provide details such as OS, Python version, and any relevant software or dependencies."
+      placeholder: e.g., macOS 13.5, Python 3.10, torch==2.3.0, Gradio 4.44.1
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: Steps to Reproduce
+      description: |
+        Include detailed steps, screenshots, and logs. Use the correct markdown syntax for code blocks.
+      placeholder: |
+        1. Create a new conda environment.
+        2. Clone the repository and install as pip package.
+        3. Run the command: `f5-tts_infer-gradio` with no ref_text provided.
+        4. Stuck there with the following message... (attach logs and also error msg e.g. after ctrl-c).
+    validations:
+      required: true
+  - type: textarea
+    attributes:
+      label: ✔️ Expected Behavior
+      placeholder: Describe what you expected to happen, e.g. output a generated audio
+    validations:
+      required: false
+  - type: textarea
+    attributes:
+      label: ❌ Actual Behavior
+      placeholder: Describe what actually happened, failure messages, etc.
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/question.yml

@@ -0,0 +1,26 @@
+name: "Question"
+description: |
+  Pure question or inquiry about the project, usage issue goes with "help wanted".
+labels:
+  - question
+body:
+  - type: checkboxes
+    attributes:
+      label: Checks
+      description: "To help us grasp quickly, please confirm the following:"
+      options:
+        - label: This template is only for question, not feature requests or bug reports.
+          required: true
+        - label: I have thoroughly reviewed the project documentation and read the related paper(s).
+          required: true
+        - label: I have searched for existing issues, including closed ones, no similar questions.
+          required: true
+        - label: I confirm that I am using English to submit this report in order to facilitate communication.
+          required: true
+  - type: textarea
+    attributes:
+      label: Question details
+      description: |
+        Question details, clearly stated using proper markdown syntax.
+    validations:
+      required: true

src/f5_tts/infer/README.md

@@ -113,4 +113,77 @@ To test speech editing capabilities, use the following command:
 
 ```bash
 python src/f5_tts/infer/speech_edit.py
-```
+```
+
+## Socket Realtime Client
+
+To communicate with socket server you need to run 
+```bash
+python src/f5_tts/socket_server.py
+```
+
+<details>
+<summary>Then create client to communicate</summary>
+
+``` python
+import socket
+import numpy as np
+import asyncio
+import pyaudio
+
+async def listen_to_voice(text, server_ip='localhost', server_port=9999):
+    client_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+    client_socket.connect((server_ip, server_port))
+
+    async def play_audio_stream():
+        buffer = b''
+        p = pyaudio.PyAudio()
+        stream = p.open(format=pyaudio.paFloat32,
+                        channels=1,
+                        rate=24000,  # Ensure this matches the server's sampling rate
+                        output=True,
+                        frames_per_buffer=2048)
+
+        try:
+            while True:
+                chunk = await asyncio.get_event_loop().run_in_executor(None, client_socket.recv, 1024)
+                if not chunk:  # End of stream
+                    break
+                if b"END_OF_AUDIO" in chunk:
+                    buffer += chunk.replace(b"END_OF_AUDIO", b"")
+                    if buffer:
+                        audio_array = np.frombuffer(buffer, dtype=np.float32).copy()  # Make a writable copy
+                        stream.write(audio_array.tobytes())
+                    break
+                buffer += chunk
+                if len(buffer) >= 4096:
+                    audio_array = np.frombuffer(buffer[:4096], dtype=np.float32).copy()  # Make a writable copy
+                    stream.write(audio_array.tobytes())
+                    buffer = buffer[4096:]
+        finally:
+            stream.stop_stream()
+            stream.close()
+            p.terminate()
+
+    try:
+        # Send only the text to the server
+        await asyncio.get_event_loop().run_in_executor(None, client_socket.sendall, text.encode('utf-8'))
+        await play_audio_stream()
+        print("Audio playback finished.")
+
+    except Exception as e:
+        print(f"Error in listen_to_voice: {e}")
+
+    finally:
+        client_socket.close()
+
+# Example usage: Replace this with your actual server IP and port
+async def main():
+    await listen_to_voice("my name is jenny..", server_ip='localhost', server_port=9998)
+
+# Run the main async function
+asyncio.run(main())
+```
+
+</details>
+

src/f5_tts/infer/utils_infer.py

@@ -218,6 +218,22 @@ def load_model(
     return model
 
 
+def remove_silence_edges(audio, silence_threshold=-42):
+    # Remove silence from the start
+    non_silent_start_idx = silence.detect_leading_silence(audio, silence_threshold=silence_threshold)
+    audio = audio[non_silent_start_idx:]
+
+    # Remove silence from the end
+    non_silent_end_duration = audio.duration_seconds
+    for ms in reversed(audio):
+        if ms.dBFS > silence_threshold:
+            break
+        non_silent_end_duration -= 0.001
+    trimmed_audio = audio[: int(non_silent_end_duration * 1000)]
+
+    return trimmed_audio
+
+
 # preprocess reference audio and text
 
 
@@ -229,7 +245,7 @@ def preprocess_ref_audio_text(ref_audio_orig, ref_text, clip_short=True, show_in
         if clip_short:
             # 1. try to find long silence for clipping
             non_silent_segs = silence.split_on_silence(
-                aseg, min_silence_len=1000, silence_thresh=-50, keep_silence=1000
+                aseg, min_silence_len=1000, silence_thresh=-50, keep_silence=1000, seek_step=10
             )
             non_silent_wave = AudioSegment.silent(duration=0)
             for non_silent_seg in non_silent_segs:
@@ -241,7 +257,7 @@ def preprocess_ref_audio_text(ref_audio_orig, ref_text, clip_short=True, show_in
             # 2. try to find short silence for clipping if 1. failed
             if len(non_silent_wave) > 15000:
                 non_silent_segs = silence.split_on_silence(
-                    aseg, min_silence_len=100, silence_thresh=-40, keep_silence=1000
+                    aseg, min_silence_len=100, silence_thresh=-40, keep_silence=1000, seek_step=10
                 )
                 non_silent_wave = AudioSegment.silent(duration=0)
                 for non_silent_seg in non_silent_segs:
@@ -257,6 +273,7 @@ def preprocess_ref_audio_text(ref_audio_orig, ref_text, clip_short=True, show_in
                 aseg = aseg[:15000]
                 show_info("Audio is over 15s, clipping short. (3)")
 
+        aseg = remove_silence_edges(aseg) + AudioSegment.silent(duration=50)
         aseg.export(f.name, format="wav")
         ref_audio = f.name
 
@@ -473,7 +490,9 @@ def infer_batch_process(
 
 def remove_silence_for_generated_wav(filename):
     aseg = AudioSegment.from_file(filename)
-    non_silent_segs = silence.split_on_silence(aseg, min_silence_len=1000, silence_thresh=-50, keep_silence=500)
+    non_silent_segs = silence.split_on_silence(
+        aseg, min_silence_len=1000, silence_thresh=-50, keep_silence=500, seek_step=10
+    )
     non_silent_wave = AudioSegment.silent(duration=0)
     for non_silent_seg in non_silent_segs:
         non_silent_wave += non_silent_seg

src/f5_tts/socket_server.py

@@ -0,0 +1,163 @@
+import socket
+import struct
+import torch
+import torchaudio
+from threading import Thread
+
+
+import gc
+import traceback
+
+
+from infer.utils_infer import infer_batch_process, preprocess_ref_audio_text, load_vocoder, load_model
+from model.backbones.dit import DiT
+
+
+class TTSStreamingProcessor:
+    def __init__(self, ckpt_file, vocab_file, ref_audio, ref_text, device=None, dtype=torch.float32):
+        self.device = device or ("cuda" if torch.cuda.is_available() else "cpu")
+
+        # Load the model using the provided checkpoint and vocab files
+        self.model = load_model(
+            model_cls=DiT,
+            model_cfg=dict(dim=1024, depth=22, heads=16, ff_mult=2, text_dim=512, conv_layers=4),
+            ckpt_path=ckpt_file,
+            mel_spec_type="vocos",  # or "bigvgan" depending on vocoder
+            vocab_file=vocab_file,
+            ode_method="euler",
+            use_ema=True,
+            device=self.device,
+        ).to(self.device, dtype=dtype)
+
+        # Load the vocoder
+        self.vocoder = load_vocoder(is_local=False)
+
+        # Set sampling rate for streaming
+        self.sampling_rate = 24000  # Consistency with client
+
+        # Set reference audio and text
+        self.ref_audio = ref_audio
+        self.ref_text = ref_text
+
+        # Warm up the model
+        self._warm_up()
+
+    def _warm_up(self):
+        """Warm up the model with a dummy input to ensure it's ready for real-time processing."""
+        print("Warming up the model...")
+        ref_audio, ref_text = preprocess_ref_audio_text(self.ref_audio, self.ref_text)
+        audio, sr = torchaudio.load(ref_audio)
+        gen_text = "Warm-up text for the model."
+
+        # Pass the vocoder as an argument here
+        infer_batch_process((audio, sr), ref_text, [gen_text], self.model, self.vocoder, device=self.device)
+        print("Warm-up completed.")
+
+    def generate_stream(self, text, play_steps_in_s=0.5):
+        """Generate audio in chunks and yield them in real-time."""
+        # Preprocess the reference audio and text
+        ref_audio, ref_text = preprocess_ref_audio_text(self.ref_audio, self.ref_text)
+
+        # Load reference audio
+        audio, sr = torchaudio.load(ref_audio)
+
+        # Run inference for the input text
+        audio_chunk, final_sample_rate, _ = infer_batch_process(
+            (audio, sr),
+            ref_text,
+            [text],
+            self.model,
+            self.vocoder,
+            device=self.device,  # Pass vocoder here
+        )
+
+        # Break the generated audio into chunks and send them
+        chunk_size = int(final_sample_rate * play_steps_in_s)
+
+        for i in range(0, len(audio_chunk), chunk_size):
+            chunk = audio_chunk[i : i + chunk_size]
+
+            # Check if it's the final chunk
+            if i + chunk_size >= len(audio_chunk):
+                chunk = audio_chunk[i:]
+
+            # Avoid sending empty or repeated chunks
+            if len(chunk) == 0:
+                break
+
+            # Pack and send the audio chunk
+            packed_audio = struct.pack(f"{len(chunk)}f", *chunk)
+            yield packed_audio
+
+        # Ensure that no final word is repeated by not resending partial chunks
+        if len(audio_chunk) % chunk_size != 0:
+            remaining_chunk = audio_chunk[-(len(audio_chunk) % chunk_size) :]
+            packed_audio = struct.pack(f"{len(remaining_chunk)}f", *remaining_chunk)
+            yield packed_audio
+
+
+def handle_client(client_socket, processor):
+    try:
+        while True:
+            # Receive data from the client
+            data = client_socket.recv(1024).decode("utf-8")
+            if not data:
+                break
+
+            try:
+                # The client sends the text input
+                text = data.strip()
+
+                # Generate and stream audio chunks
+                for audio_chunk in processor.generate_stream(text):
+                    client_socket.sendall(audio_chunk)
+
+                # Send end-of-audio signal
+                client_socket.sendall(b"END_OF_AUDIO")
+
+            except Exception as inner_e:
+                print(f"Error during processing: {inner_e}")
+                traceback.print_exc()  # Print the full traceback to diagnose the issue
+                break
+
+    except Exception as e:
+        print(f"Error handling client: {e}")
+        traceback.print_exc()
+    finally:
+        client_socket.close()
+
+
+def start_server(host, port, processor):
+    server = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+    server.bind((host, port))
+    server.listen(5)
+    print(f"Server listening on {host}:{port}")
+
+    while True:
+        client_socket, addr = server.accept()
+        print(f"Accepted connection from {addr}")
+        client_handler = Thread(target=handle_client, args=(client_socket, processor))
+        client_handler.start()
+
+
+if __name__ == "__main__":
+    try:
+        # Load the model and vocoder using the provided files
+        ckpt_file = ""  # pointing your checkpoint "ckpts/model/model_1096.pt"
+        vocab_file = ""  # Add vocab file path if needed
+        ref_audio = ""  # add ref audio"./tests/ref_audio/reference.wav"
+        ref_text = ""
+
+        # Initialize the processor with the model and vocoder
+        processor = TTSStreamingProcessor(
+            ckpt_file=ckpt_file,
+            vocab_file=vocab_file,
+            ref_audio=ref_audio,
+            ref_text=ref_text,
+            dtype=torch.float32,
+        )
+
+        # Start the server
+        start_server("0.0.0.0", 9998, processor)
+    except KeyboardInterrupt:
+        gc.collect()

src/f5_tts/train/finetune_cli.py

@@ -55,7 +55,6 @@ def parse_args():
         default=None,
         help="Path to custom tokenizer vocab file (only used if tokenizer = 'custom')",
     )
-
     parser.add_argument(
         "--log_samples",
         type=bool,
@@ -63,6 +62,12 @@ def parse_args():
         help="Log inferenced samples per ckpt save steps",
     )
     parser.add_argument("--logger", type=str, default=None, choices=["wandb", "tensorboard"], help="logger")
+    parser.add_argument(
+        "--bnb_optimizer",
+        type=bool,
+        default=False,
+        help="Use 8-bit Adam optimizer from bitsandbytes",
+    )
 
     return parser.parse_args()
 
@@ -147,6 +152,7 @@ def main():
         wandb_resume_id=wandb_resume_id,
         log_samples=args.log_samples,
         last_per_steps=args.last_per_steps,
+        bnb_optimizer=args.bnb_optimizer,
     )
 
     train_dataset = load_dataset(args.dataset_name, tokenizer, mel_spec_kwargs=mel_spec_kwargs)

src/f5_tts/train/finetune_gradio.py

@@ -1372,7 +1372,7 @@ def get_audio_select(file_sample):
 with gr.Blocks() as app:
     gr.Markdown(
         """
-# E2/F5 TTS AUTOMATIC FINETUNE 
+# E2/F5 TTS Automatic Finetune
 
 This is a local web UI for F5 TTS with advanced batch processing support. This app supports the following TTS models:
 
@@ -1381,35 +1381,35 @@ This is a local web UI for F5 TTS with advanced batch processing support. This a
 
 The checkpoints support English and Chinese.
 
-for tutorial and updates check here (https://github.com/SWivid/F5-TTS/discussions/143)
+For tutorial and updates check here (https://github.com/SWivid/F5-TTS/discussions/143)
 """
     )
 
     with gr.Row():
         projects, projects_selelect = get_list_projects()
-        tokenizer_type = gr.Radio(label="Tokenizer Type", choices=["pinyin", "char"], value="pinyin")
-        project_name = gr.Textbox(label="project name", value="my_speak")
-        bt_create = gr.Button("create new project")
+        tokenizer_type = gr.Radio(label="Tokenizer Type", choices=["pinyin", "char", "custom"], value="pinyin")
+        project_name = gr.Textbox(label="Project Name", value="my_speak")
+        bt_create = gr.Button("Create a New Project")
 
     with gr.Row():
         cm_project = gr.Dropdown(
             choices=projects, value=projects_selelect, label="Project", allow_custom_value=True, scale=6
         )
-        ch_refresh_project = gr.Button("refresh", scale=1)
+        ch_refresh_project = gr.Button("Refresh", scale=1)
 
     bt_create.click(fn=create_data_project, inputs=[project_name, tokenizer_type], outputs=[cm_project])
 
     with gr.Tabs():
-        with gr.TabItem("transcribe Data"):
+        with gr.TabItem("Transcribe Data"):
             gr.Markdown("""```plaintext 
 Skip this step if you have your dataset, metadata.csv, and a folder wavs with all the audio files.                 
 ```""")
 
-            ch_manual = gr.Checkbox(label="audio from path", value=False)
+            ch_manual = gr.Checkbox(label="Audio from Path", value=False)
 
             mark_info_transcribe = gr.Markdown(
                 """```plaintext    
-     Place your 'wavs' folder and 'metadata.csv' file in the {your_project_name}' directory. 
+     Place your 'wavs' folder and 'metadata.csv' file in the '{your_project_name}' directory. 
                  
      my_speak/
      │
@@ -1421,10 +1421,10 @@ Skip this step if you have your dataset, metadata.csv, and a folder wavs with al
                 visible=False,
             )
 
-            audio_speaker = gr.File(label="voice", type="filepath", file_count="multiple")
-            txt_lang = gr.Text(label="Language", value="english")
-            bt_transcribe = bt_create = gr.Button("transcribe")
-            txt_info_transcribe = gr.Text(label="info", value="")
+            audio_speaker = gr.File(label="Voice", type="filepath", file_count="multiple")
+            txt_lang = gr.Text(label="Language", value="English")
+            bt_transcribe = bt_create = gr.Button("Transcribe")
+            txt_info_transcribe = gr.Text(label="Info", value="")
             bt_transcribe.click(
                 fn=transcribe_all,
                 inputs=[cm_project, audio_speaker, txt_lang, ch_manual],
@@ -1432,7 +1432,7 @@ Skip this step if you have your dataset, metadata.csv, and a folder wavs with al
             )
             ch_manual.change(fn=check_user, inputs=[ch_manual], outputs=[audio_speaker, mark_info_transcribe])
 
-            random_sample_transcribe = gr.Button("random sample")
+            random_sample_transcribe = gr.Button("Random Sample")
 
             with gr.Row():
                 random_text_transcribe = gr.Text(label="Text")
@@ -1444,16 +1444,16 @@ Skip this step if you have your dataset, metadata.csv, and a folder wavs with al
                 outputs=[random_text_transcribe, random_audio_transcribe],
             )
 
-        with gr.TabItem("vocab check"):
+        with gr.TabItem("Vocab Check"):
             gr.Markdown("""```plaintext 
-check the vocabulary for fine-tuning Emilia_ZH_EN to ensure all symbols are included. for finetune new language
+Check the vocabulary for fine-tuning Emilia_ZH_EN to ensure all symbols are included. For fine-tuning a new language.
 ```""")
 
-            check_button = gr.Button("check vocab")
-            txt_info_check = gr.Text(label="info", value="")
+            check_button = gr.Button("Check Vocab")
+            txt_info_check = gr.Text(label="Info", value="")
 
             gr.Markdown("""```plaintext 
-Using the extended model, you can fine-tune to a new language that is missing symbols in the vocab , this create a new model with a new vocabulary size and save it in your ckpts/project folder.
+Using the extended model, you can finetune to a new language that is missing symbols in the vocab. This creates a new model with a new vocabulary size and saves it in your ckpts/project folder.
 ```""")
 
             exp_name_extend = gr.Radio(label="Model", choices=["F5-TTS", "E2-TTS"], value="F5-TTS")
@@ -1465,10 +1465,10 @@ Using the extended model, you can fine-tune to a new language that is missing sy
                     placeholder="To add new symbols, make sure to use ',' for each symbol",
                     scale=6,
                 )
-                txt_count_symbol = gr.Textbox(label="new size vocab", value="", scale=1)
+                txt_count_symbol = gr.Textbox(label="New Vocab Size", value="", scale=1)
 
-            extend_button = gr.Button("Extended")
-            txt_info_extend = gr.Text(label="info", value="")
+            extend_button = gr.Button("Extend")
+            txt_info_extend = gr.Text(label="Info", value="")
 
             txt_extend.change(vocab_count, inputs=[txt_extend], outputs=[txt_count_symbol])
             check_button.click(fn=vocab_check, inputs=[cm_project], outputs=[txt_info_check, txt_extend])
@@ -1476,18 +1476,18 @@ Using the extended model, you can fine-tune to a new language that is missing sy
                 fn=vocab_extend, inputs=[cm_project, txt_extend, exp_name_extend], outputs=[txt_info_extend]
             )
 
-        with gr.TabItem("prepare Data"):
+        with gr.TabItem("Prepare Data"):
             gr.Markdown("""```plaintext 
-Skip this step if you have your dataset, raw.arrow , duraction.json and vocab.txt
+Skip this step if you have your dataset, raw.arrow, duration.json, and vocab.txt
 ```""")
 
             gr.Markdown(
                 """```plaintext    
-     place all your wavs folder and your metadata.csv file in {your name project}  
+     Place all your "wavs" folder and your "metadata.csv" file in your project name directory.
 
-     suport format for audio "wav", "mp3", "aac", "flac", "m4a", "alac", "ogg", "aiff", "wma", "amr"
+     Supported audio formats: "wav", "mp3", "aac", "flac", "m4a", "alac", "ogg", "aiff", "wma", "amr"
 
-     example wav format                               
+     Example wav format:                               
      my_speak/
      │
      ├── wavs/
@@ -1497,24 +1497,24 @@ Skip this step if you have your dataset, raw.arrow , duraction.json and vocab.tx
      │
      └── metadata.csv
       
-     file format metadata.csv 
+     File format metadata.csv:
 
      audio1|text1 or audio1.wav|text1 or your_path/audio1.wav|text1 
-     audio2|text1 or audio2.wav|text1 or your_path/audio1.wav|text1 
+     audio2|text1 or audio2.wav|text1 or your_path/audio2.wav|text1 
      ...
 
      ```"""
             )
-            ch_tokenizern = gr.Checkbox(label="create vocabulary", value=False, visible=False)
-            bt_prepare = bt_create = gr.Button("prepare")
-            txt_info_prepare = gr.Text(label="info", value="")
-            txt_vocab_prepare = gr.Text(label="vocab", value="")
+            ch_tokenizern = gr.Checkbox(label="Create Vocabulary", value=False, visible=False)
+            bt_prepare = bt_create = gr.Button("Prepare")
+            txt_info_prepare = gr.Text(label="Info", value="")
+            txt_vocab_prepare = gr.Text(label="Vocab", value="")
 
             bt_prepare.click(
                 fn=create_metadata, inputs=[cm_project, ch_tokenizern], outputs=[txt_info_prepare, txt_vocab_prepare]
             )
 
-            random_sample_prepare = gr.Button("random sample")
+            random_sample_prepare = gr.Button("Random Sample")
 
             with gr.Row():
                 random_text_prepare = gr.Text(label="Tokenizer")
@@ -1524,20 +1524,20 @@ Skip this step if you have your dataset, raw.arrow , duraction.json and vocab.tx
                 fn=get_random_sample_prepare, inputs=[cm_project], outputs=[random_text_prepare, random_audio_prepare]
             )
 
-        with gr.TabItem("train Data"):
+        with gr.TabItem("Train Data"):
             gr.Markdown("""```plaintext 
-The auto-setting is still experimental. Please make sure that the epochs , save per updates , and last per steps are set correctly, or change them manually as needed.
+The auto-setting is still experimental. Please make sure that the epochs, save per updates, and last per steps are set correctly, or change them manually as needed.
 If you encounter a memory error, try reducing the batch size per GPU to a smaller number.
 ```""")
             with gr.Row():
                 bt_calculate = bt_create = gr.Button("Auto Settings")
-                lb_samples = gr.Label(label="samples")
+                lb_samples = gr.Label(label="Samples")
                 batch_size_type = gr.Radio(label="Batch Size Type", choices=["frame", "sample"], value="frame")
 
             with gr.Row():
-                ch_finetune = bt_create = gr.Checkbox(label="finetune", value=True)
+                ch_finetune = bt_create = gr.Checkbox(label="Finetune", value=True)
                 tokenizer_file = gr.Textbox(label="Tokenizer File", value="")
-                file_checkpoint_train = gr.Textbox(label="Path to the preetrain checkpoint ", value="")
+                file_checkpoint_train = gr.Textbox(label="Path to the Pretrained Checkpoint", value="")
 
             with gr.Row():
                 exp_name = gr.Radio(label="Model", choices=["F5TTS_Base", "E2TTS_Base"], value="F5TTS_Base")
@@ -1603,8 +1603,8 @@ If you encounter a memory error, try reducing the batch size per GPU to a smalle
                 mixed_precision.value = mixed_precisionv
                 cd_logger.value = cd_loggerv
 
-            ch_stream = gr.Checkbox(label="stream output experiment.", value=True)
-            txt_info_train = gr.Text(label="info", value="")
+            ch_stream = gr.Checkbox(label="Stream Output Experiment", value=True)
+            txt_info_train = gr.Text(label="Info", value="")
 
             list_audios, select_audio = get_audio_project(projects_selelect, False)
 
@@ -1619,18 +1619,18 @@ If you encounter a memory error, try reducing the batch size per GPU to a smalle
                 ch_list_audio = gr.Dropdown(
                     choices=list_audios,
                     value=select_audio,
-                    label="audios",
+                    label="Audios",
                     allow_custom_value=True,
                     scale=6,
                     interactive=True,
                 )
-                bt_stream_audio = gr.Button("refresh", scale=1)
+                bt_stream_audio = gr.Button("Refresh", scale=1)
                 bt_stream_audio.click(fn=get_audio_project, inputs=[cm_project], outputs=[ch_list_audio])
                 cm_project.change(fn=get_audio_project, inputs=[cm_project], outputs=[ch_list_audio])
 
             with gr.Row():
-                audio_ref_stream = gr.Audio(label="original", type="filepath", value=select_audio_ref)
-                audio_gen_stream = gr.Audio(label="generate", type="filepath", value=select_audio_gen)
+                audio_ref_stream = gr.Audio(label="Original", type="filepath", value=select_audio_ref)
+                audio_gen_stream = gr.Audio(label="Generate", type="filepath", value=select_audio_gen)
 
             ch_list_audio.change(
                 fn=get_audio_select,
@@ -1730,36 +1730,36 @@ If you encounter a memory error, try reducing the batch size per GPU to a smalle
                 outputs=outputs,
             )
 
-        with gr.TabItem("test model"):
+        with gr.TabItem("Test Model"):
             gr.Markdown("""```plaintext 
-SOS : check the use_ema setting (True or False) for your model to see what works best for you. 
+SOS: Check the use_ema setting (True or False) for your model to see what works best for you. 
 ```""")
             exp_name = gr.Radio(label="Model", choices=["F5-TTS", "E2-TTS"], value="F5-TTS")
             list_checkpoints, checkpoint_select = get_checkpoints_project(projects_selelect, False)
 
-            nfe_step = gr.Number(label="n_step", value=32)
-            ch_use_ema = gr.Checkbox(label="use ema", value=True)
+            nfe_step = gr.Number(label="NFE Step", value=32)
+            ch_use_ema = gr.Checkbox(label="Use EMA", value=True)
             with gr.Row():
                 cm_checkpoint = gr.Dropdown(
-                    choices=list_checkpoints, value=checkpoint_select, label="checkpoints", allow_custom_value=True
+                    choices=list_checkpoints, value=checkpoint_select, label="Checkpoints", allow_custom_value=True
                 )
-                bt_checkpoint_refresh = gr.Button("refresh")
+                bt_checkpoint_refresh = gr.Button("Refresh")
 
-            random_sample_infer = gr.Button("random sample")
+            random_sample_infer = gr.Button("Random Sample")
 
-            ref_text = gr.Textbox(label="ref text")
-            ref_audio = gr.Audio(label="audio ref", type="filepath")
-            gen_text = gr.Textbox(label="gen text")
+            ref_text = gr.Textbox(label="Ref Text")
+            ref_audio = gr.Audio(label="Audio Ref", type="filepath")
+            gen_text = gr.Textbox(label="Gen Text")
 
             random_sample_infer.click(
                 fn=get_random_sample_infer, inputs=[cm_project], outputs=[ref_text, gen_text, ref_audio]
             )
 
             with gr.Row():
-                txt_info_gpu = gr.Textbox("", label="device")
-                check_button_infer = gr.Button("infer")
+                txt_info_gpu = gr.Textbox("", label="Device")
+                check_button_infer = gr.Button("Infer")
 
-            gen_audio = gr.Audio(label="audio gen", type="filepath")
+            gen_audio = gr.Audio(label="Audio Gen", type="filepath")
 
             check_button_infer.click(
                 fn=infer,
@@ -1770,22 +1770,22 @@ SOS : check the use_ema setting (True or False) for your model to see what works
             bt_checkpoint_refresh.click(fn=get_checkpoints_project, inputs=[cm_project], outputs=[cm_checkpoint])
             cm_project.change(fn=get_checkpoints_project, inputs=[cm_project], outputs=[cm_checkpoint])
 
-        with gr.TabItem("reduse checkpoint"):
+        with gr.TabItem("Reduce Checkpoint"):
             gr.Markdown("""```plaintext 
-Reduce the model size from 5GB to 1.3GB. The new checkpoint can be used for inference or fine-tuning afterward, but it cannot be used to continue training..
+Reduce the model size from 5GB to 1.3GB. The new checkpoint can be used for inference or fine-tuning afterward, but it cannot be used to continue training.
 ```""")
-            txt_path_checkpoint = gr.Text(label="path checkpoint :")
-            txt_path_checkpoint_small = gr.Text(label="path output :")
-            ch_safetensors = gr.Checkbox(label="safetensors", value="")
-            txt_info_reduse = gr.Text(label="info", value="")
-            reduse_button = gr.Button("reduse")
+            txt_path_checkpoint = gr.Text(label="Path to Checkpoint:")
+            txt_path_checkpoint_small = gr.Text(label="Path to Output:")
+            ch_safetensors = gr.Checkbox(label="Safetensors", value="")
+            txt_info_reduse = gr.Text(label="Info", value="")
+            reduse_button = gr.Button("Reduce")
             reduse_button.click(
                 fn=extract_and_save_ema_model,
                 inputs=[txt_path_checkpoint, txt_path_checkpoint_small, ch_safetensors],
                 outputs=[txt_info_reduse],
             )
 
-        with gr.TabItem("system info"):
+        with gr.TabItem("System Info"):
             output_box = gr.Textbox(label="GPU and CPU Information", lines=20)
 
             def update_stats():