Update app.py

app.py

@@ -14,11 +14,6 @@ from pydub import AudioSegment
 
 final_audio = AudioSegment.empty()
 
-from pydub import AudioSegment
-from bark import generate_audio  # Importing Bark
-
-
-
 
 # Voice model
 VOICE_MODEL = "tts_models/en/vctk/vits"
@@ -116,98 +111,88 @@ SPEAKER_METADATA = {
 	    273: { "age": 18, "gender": "F", "accent": "English"}
 	}
 
+def clean_text(text):
+    # Remove hyperlinks
+    return re.sub(r'http[s]?://\S+', '', text)
 
+def extract_paragraphs_from_docx(docx_file):
+    document = Document(docx_file.name)
+    paragraphs = [p.text.strip() for p in document.paragraphs if p.text.strip()]
+    return [clean_text(p) for p in paragraphs]
 
-# Bark prompts (example)
-BARK_PROMPTS = [
-    "Shy girl",
-    "Old man",
-    "Excited child",
-    "Angry woman"
-]
-
-def list_speaker_choices(metadata):
-    """Helper function to list speakers from metadata (for VCTK and Coqui)"""
-    return [f"Speaker {sid} | {meta['gender']} | {meta['accent']}" for sid, meta in SPEAKER_METADATA.items()]
+def list_speaker_choices():
+    return [f"{sid} | {meta['gender']} | {meta['accent']}" for sid, meta in SPEAKER_METADATA.items()]
 
 def get_speaker_id_from_label(label):
-    """Extract speaker ID from label string"""
     return label.split('|')[0].strip()
 
-def generate_audio(sample_text, speaker_label, engine):
-    """Generate audio based on engine choice"""
+def generate_sample_audio(sample_text, speaker_label):
+    if len(sample_text) > 500:
+        raise gr.Error("Sample text exceeds 500 characters.")
     speaker_id = get_speaker_id_from_label(speaker_label)
-    model = None
-
-    # Engine selection logic
-    if engine == "bark":
-        model = TTS("bark_model_path")  # Replace with actual path for Bark model
-    elif engine == "coqui":
-        model = TTS("tts_models/multilingual/multi-dataset/xtts_v2")  # Replace with actual path for Coqui model
-    elif engine == "vctk":
-        model = TTS(VOICE_MODEL)  # Replace with actual path for VCTK model
-
-    # Temporary file creation for output audio
+    model = TTS("tts_models/en/vctk/vits")
     with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp_wav:
         model.tts_to_file(text=sample_text, speaker="p"+speaker_id, file_path=tmp_wav.name)
         return tmp_wav.name
 
-# --- UI Components ---
+def generate_audio(docx_file, speaker_label):
+    speaker_id = get_speaker_id_from_label(speaker_label)
+    model = TTS("tts_models/en/vctk/vits")
+
+    paragraphs = extract_paragraphs_from_docx(docx_file)
+    combined_audio = AudioSegment.empty()
+    temp_files = []
+
+    try:
+        for idx, para in enumerate(paragraphs):
+            tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
+            model.tts_to_file(text=para, speaker="p"+speaker_id, file_path=tmp.name)
+            audio_chunk = AudioSegment.from_wav(tmp.name)
+            combined_audio += audio_chunk
+            temp_files.append(tmp.name)
+            tmp.close()
+
+    except Exception as e:
+        print("Generation interrupted. Saving partial output.", e)
+
+    output_dir = tempfile.mkdtemp()
+    final_output_path = os.path.join(output_dir, "final_output.wav")
+    combined_audio.export(final_output_path, format="wav")
+
+    zip_path = os.path.join(output_dir, "output.zip")
+    with zipfile.ZipFile(zip_path, 'w') as zipf:
+        zipf.write(final_output_path, arcname="final_output.wav")
+
+    for f in temp_files:
+        os.remove(f)
+
+    return zip_path
+
+# --- UI ---
+speaker_choices = list_speaker_choices()
+
 with gr.Blocks() as demo:
-    gr.Markdown("## 📄 TTS Voice Generator with Multiple Engines")
-    
-    # Engine dropdown
-    engine_dropdown = gr.Dropdown(
-        label="Select TTS Engine", 
-        choices=["bark", "coqui", "vctk"], 
-        value="vctk"
-    )
-
-    # Speaker/Prompt dropdown (dynamic)
-    speaker_dropdown = gr.Dropdown(label="Select Speaker", visible=False)
-    prompt_dropdown = gr.Dropdown(label="Select Prompt", visible=False)
-
-    # Sample text box
-    sample_textbox = gr.Textbox(label="Enter Sample Text (Max 500 characters)", max_lines=5)
-    sample_audio = gr.Audio(label="Sample Output Audio", type="filepath")
-    
-    # Define metadata choices for speakers (Coqui and VCTK)
-    speaker_choices_vctk_coqui = list_speaker_choices(SPEAKER_METADATA)
-    speaker_dropdown.choices = speaker_choices_vctk_coqui  # Use metadata for VCTK/Coqui speakers
-
-    # Define Bark prompts (choose from predefined prompts)
-    prompt_dropdown.choices = BARK_PROMPTS
-    
-    # Dynamically update dropdown visibility based on engine selection
-    def update_dropdowns(engine):
-        if engine == "bark":
-            speaker_dropdown.visible = False
-            prompt_dropdown.visible = True
-        elif engine == "coqui" or engine == "vctk":
-            speaker_dropdown.visible = True
-            prompt_dropdown.visible = False
-        return gr.update(visible=speaker_dropdown.visible), gr.update(visible=prompt_dropdown.visible)
-
-    # Trigger dropdown visibility changes
-    engine_dropdown.change(update_dropdowns, inputs=engine_dropdown, outputs=[speaker_dropdown, prompt_dropdown])
-
-    # Button to generate audio from sample text
-    generate_button = gr.Button("Generate Audio")
-    generate_button.click(
-        fn=generate_audio, 
-        inputs=[sample_textbox, speaker_dropdown, engine_dropdown], 
-        outputs=[sample_audio]
-    )
-
-    # Button to clear the sample text and audio
-    def clear_sample():
-        return "", None
-
-    clear_button = gr.Button("Clear")
-    clear_button.click(fn=clear_sample, inputs=[], outputs=[sample_textbox, sample_audio])
+    gr.Markdown("## 📄 TTS Voice Generator with Paragraph-Wise Processing")
 
+    with gr.Row():
+        speaker_dropdown = gr.Dropdown(label="Select Voice", choices=speaker_choices)
 
-if __name__ == "__main__":
-    demo.launch()
+    with gr.Row():
+        sample_textbox = gr.Textbox(label="Enter Sample Text (Max 500 characters)", max_lines=5)
+        sample_button = gr.Button("Generate Sample")
+        clear_button = gr.Button("Clear Sample")
+
+    sample_audio = gr.Audio(label="Sample Output", type="filepath")
+
+    sample_button.click(fn=generate_sample_audio, inputs=[sample_textbox, speaker_dropdown], outputs=[sample_audio])
+    clear_button.click(fn=lambda: None, inputs=[], outputs=[sample_audio])
 
+    with gr.Row():
+        docx_input = gr.File(label="Upload DOCX File", file_types=[".docx"])
+        generate_button = gr.Button("Generate Full Audio")
+        download_output = gr.File(label="Download Output Zip")
 
+    generate_button.click(fn=generate_audio, inputs=[docx_input, speaker_dropdown], outputs=[download_output])
+
+if __name__ == "__main__":
+    demo.launch()