Update app.py

app.py

@@ -1,4 +1,6 @@
 import os
+import subprocess
+import tempfile
 
 import gradio as gr
 import numpy as np
@@ -105,6 +107,47 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 generator = load_csm_1b(device=device)
 
 
+def convert_ebook_to_txt(ebook_path):
+    """Convert an ebook file to text using Calibre's ebook-convert."""
+    if not ebook_path:
+        return None
+    
+    # Create a temporary file for the output
+    with tempfile.NamedTemporaryFile(suffix='.txt', delete=False) as temp_txt:
+        txt_path = temp_txt.name
+    
+    try:
+        # Run ebook-convert from Calibre
+        subprocess.run(
+            ["ebook-convert", ebook_path, txt_path], 
+            check=True, 
+            stdout=subprocess.PIPE, 
+            stderr=subprocess.PIPE
+        )
+        
+        # Read the converted text
+        with open(txt_path, 'r', encoding='utf-8') as f:
+            text_content = f.read()
+        
+        # Clean up
+        os.unlink(txt_path)
+        
+        # Format the text into alternating lines for conversation
+        lines = [line.strip() for line in text_content.split('.') if line.strip()]
+        formatted_lines = []
+        
+        # Take up to 20 sentences to avoid extremely long conversations
+        for i, line in enumerate(lines[:20]):
+            formatted_lines.append(line + ".")
+        
+        return "\n".join(formatted_lines)
+    
+    except Exception as e:
+        if os.path.exists(txt_path):
+            os.unlink(txt_path)
+        raise gr.Error(f"Error converting ebook: {str(e)}")
+
+
 @spaces.GPU(duration=gpu_timeout)
 def infer(
     text_prompt_speaker_a,
@@ -203,6 +246,20 @@ def create_speaker_prompt_ui(speaker_name: str):
     return speaker_dropdown, text_prompt_speaker, audio_prompt_speaker
 
 
+def process_ebook(ebook_file):
+    if ebook_file is None:
+        return None
+    text_content = convert_ebook_to_txt(ebook_file)
+    return text_content
+
+
+def update_input_method(choice):
+    if choice == "text_input":
+        return gr.update(visible=True), gr.update(visible=False), None
+    else:
+        return gr.update(visible=False), gr.update(visible=True), None
+
+
 with gr.Blocks() as app:
     gr.Markdown(SPACE_INTRO_TEXT)
     gr.Markdown("## Voices")
@@ -236,8 +293,35 @@ with gr.Blocks() as app:
     speaker_b_dropdown.change(fn=update_text, inputs=[speaker_b_dropdown], outputs=[text_prompt_speaker_b])
 
     gr.Markdown(CONVO_INTRO_TEXT)
-
-    gen_conversation_input = gr.TextArea(label="conversation", lines=20, value=DEFAULT_CONVERSATION)
+    
+    # Radio button for selecting input method
+    input_method = gr.Radio(
+        ["Direct text input", "Upload ebook file"], 
+        label="Choose input method", 
+        value="Direct text input"
+    )
+    
+    # Container for text input method
+    with gr.Group(visible=True) as text_input_group:
+        gen_conversation_input = gr.TextArea(label="Conversation", lines=20, value=DEFAULT_CONVERSATION)
+    
+    # Container for ebook upload method
+    with gr.Group(visible=False) as ebook_input_group:
+        ebook_file = gr.File(label="Upload ebook file (will be converted using Calibre)", file_types=[".epub", ".mobi", ".azw", ".azw3", ".fb2", ".pdf"])
+        process_ebook_btn = gr.Button("Process Ebook")
+    
+    input_method.change(
+        fn=lambda choice: update_input_method("text_input" if choice == "Direct text input" else "ebook"),
+        inputs=[input_method],
+        outputs=[text_input_group, ebook_input_group, gen_conversation_input]
+    )
+    
+    process_ebook_btn.click(
+        fn=process_ebook,
+        inputs=[ebook_file],
+        outputs=[gen_conversation_input]
+    )
+    
     generate_btn = gr.Button("Generate conversation", variant="primary")
     gr.Markdown("GPU time limited to 3 minutes, for longer usage duplicate the space.")
     audio_output = gr.Audio(label="Synthesized audio")
@@ -254,4 +338,4 @@ with gr.Blocks() as app:
         outputs=[audio_output],
     )
 
-app.launch(ssr_mode=True)
+app.launch(ssr_mode=True)