# FILE: app.py import gradio as gr import os import asyncio from openai import AsyncOpenAI from functools import partial import datetime # Remove create_examples_ui from ui_layout imports if it's not used elsewhere from ui_layout import ( create_main_input_components, create_speaker_config_components, create_action_and_output_components, # Removed create_examples_ui TTS_MODELS_AVAILABLE, MODEL_DEFAULT_ENV, APP_AVAILABLE_VOICES, DEFAULT_GLOBAL_VOICE, VIBE_CHOICES, DEFAULT_VIBE, PREDEFINED_VIBES ) from event_handlers import ( handle_script_processing, handle_calculate_cost, handle_speaker_config_method_visibility_change, handle_tts_model_change, handle_load_refresh_per_speaker_ui_trigger, handle_dynamic_accordion_input_change, get_speakers_from_script ) # --- Secrets and Client Setup (Same as before) --- OPENAI_API_KEY = os.getenv("OPENAI_API_KEY") NSFW_API_URL_TEMPLATE = os.getenv("NSFW_API_URL_TEMPLATE") MODEL_DEFAULT_FROM_ENV = os.getenv("MODEL_DEFAULT", MODEL_DEFAULT_ENV) EFFECTIVE_MODEL_DEFAULT = MODEL_DEFAULT_FROM_ENV if MODEL_DEFAULT_FROM_ENV in TTS_MODELS_AVAILABLE else MODEL_DEFAULT_ENV async_openai_client = None if not OPENAI_API_KEY: # ... (secret loading logic) ... pass if OPENAI_API_KEY: async_openai_client = AsyncOpenAI(api_key=OPENAI_API_KEY) else: print("CRITICAL ERROR: OPENAI_API_KEY secret is not set.") # --- Main Blocks UI Definition --- with gr.Blocks(theme=gr.themes.Soft(), elem_id="main_blocks_ui") as demo: gr.Markdown("# Dialogue Script to Speech (OpenAI TTS) 💪💪💪 TTS = Teachers Together Strong 💪💪💪") if not OPENAI_API_KEY or not async_openai_client: gr.Markdown("

⚠️ Warning: OPENAI_API_KEY not set or invalid. Audio generation will fail.

") speaker_configs_state = gr.State({}) # --- Create Main UI Components --- (script_input, tts_model_dropdown, pause_input, global_speed_input, global_instructions_input) = create_main_input_components(EFFECTIVE_MODEL_DEFAULT) (speaker_config_method_dropdown, single_voice_group, global_voice_dropdown, detailed_per_speaker_ui_group_container, load_per_speaker_ui_button) = create_speaker_config_components() (calculate_cost_button, generate_button, cost_output, individual_lines_zip_output, merged_dialogue_mp3_output, status_output) = create_action_and_output_components() # --- Dynamic UI (@gr.render) Definition (Same as before) --- with detailed_per_speaker_ui_group_container: @gr.render( inputs=[script_input, speaker_configs_state, tts_model_dropdown], triggers=[load_per_speaker_ui_button.click, tts_model_dropdown.change] ) def render_dynamic_speaker_ui(current_script_text: str, current_speaker_configs: dict, current_tts_model: str): # ... (Full @gr.render implementation from previous correct step) ... print(f"DEBUG: @gr.render CALLED. Model: {current_tts_model}. Script: '{current_script_text[:30]}...'. State Keys: {list(current_speaker_configs.keys()) if isinstance(current_speaker_configs,dict) else 'Not a dict'}") unique_speakers = get_speakers_from_script(current_script_text) if not unique_speakers: gr.Markdown("

Enter script & click 'Load/Refresh' for per-speaker settings.

") return for speaker_idx, speaker_name in enumerate(unique_speakers): if not isinstance(current_speaker_configs, dict): current_speaker_configs = {} speaker_specific_config = current_speaker_configs.get(speaker_name, {}) accordion_elem_id = f"accordion_spk_{speaker_idx}_{speaker_name.replace(' ','_').lower()}" with gr.Accordion(f"Settings for Speaker: {speaker_name}", open=False, elem_id=accordion_elem_id): gr.Markdown(f"Configure voice for **{speaker_name}** using **{current_tts_model}** model.") default_voice = speaker_specific_config.get("voice", DEFAULT_GLOBAL_VOICE) voice_dd_elem_id = f"voice_dd_spk_{speaker_idx}" voice_dropdown = gr.Dropdown(APP_AVAILABLE_VOICES, value=default_voice, label="Voice", elem_id=voice_dd_elem_id) voice_dropdown.change(fn=partial(handle_dynamic_accordion_input_change, speaker_name=speaker_name, config_key="voice"), inputs=[voice_dropdown, speaker_configs_state], outputs=[speaker_configs_state]) if current_tts_model in ["tts-1", "tts-1-hd"]: default_speed = float(speaker_specific_config.get("speed", 1.0)) speed_slider_elem_id = f"speed_slider_spk_{speaker_idx}" speed_slider = gr.Slider(minimum=0.25, maximum=4.0, value=default_speed, step=0.05, label="Speed", elem_id=speed_slider_elem_id) speed_slider.change(fn=partial(handle_dynamic_accordion_input_change, speaker_name=speaker_name, config_key="speed"), inputs=[speed_slider, speaker_configs_state], outputs=[speaker_configs_state]) elif current_tts_model == "gpt-4o-mini-tts": default_vibe = speaker_specific_config.get("vibe", DEFAULT_VIBE) vibe_dd_elem_id = f"vibe_dd_spk_{speaker_idx}" vibe_dropdown = gr.Dropdown(VIBE_CHOICES, value=default_vibe, label="Vibe/Emotion", elem_id=vibe_dd_elem_id) default_custom_instructions = speaker_specific_config.get("custom_instructions", "") custom_instr_tb_elem_id = f"custom_instr_tb_spk_{speaker_idx}" custom_instructions_textbox = gr.Textbox(label="Custom Instructions", value=default_custom_instructions, placeholder="e.g., Speak slightly hesitant.", lines=2, visible=(default_vibe == "Custom..."), elem_id=custom_instr_tb_elem_id) vibe_dropdown.change(fn=partial(handle_dynamic_accordion_input_change, speaker_name=speaker_name, config_key="vibe"), inputs=[vibe_dropdown, speaker_configs_state], outputs=[speaker_configs_state]).then(fn=lambda vibe_val: gr.update(visible=(vibe_val == "Custom...")), inputs=[vibe_dropdown], outputs=[custom_instructions_textbox]) custom_instructions_textbox.change(fn=partial(handle_dynamic_accordion_input_change, speaker_name=speaker_name, config_key="custom_instructions"), inputs=[custom_instructions_textbox, speaker_configs_state], outputs=[speaker_configs_state]) # --- Event Listeners (Same as before) --- tts_model_dropdown.change(fn=handle_tts_model_change, inputs=[tts_model_dropdown, speaker_configs_state], outputs=[global_speed_input, global_instructions_input, speaker_configs_state]) speaker_config_method_dropdown.change(fn=handle_speaker_config_method_visibility_change, inputs=[speaker_config_method_dropdown], outputs=[single_voice_group, detailed_per_speaker_ui_group_container]) load_per_speaker_ui_button.click(fn=handle_load_refresh_per_speaker_ui_trigger, inputs=[script_input, speaker_configs_state, tts_model_dropdown], outputs=[speaker_configs_state]) calculate_cost_button.click(fn=handle_calculate_cost, inputs=[script_input, tts_model_dropdown], outputs=[cost_output]) generate_button_fn = partial(handle_script_processing, OPENAI_API_KEY, async_openai_client, NSFW_API_URL_TEMPLATE) generate_button.click(fn=generate_button_fn, inputs=[script_input, tts_model_dropdown, pause_input, speaker_config_method_dropdown, global_voice_dropdown, speaker_configs_state, global_speed_input, global_instructions_input], outputs=[individual_lines_zip_output, merged_dialogue_mp3_output, status_output]) # --- Examples Section Definition (Moved here) --- gr.Markdown("## Example Scripts") # Keep the header if desired # Define the lists needed for Examples right here example_inputs_list_comps = [ script_input, tts_model_dropdown, pause_input, speaker_config_method_dropdown, global_voice_dropdown, global_speed_input, global_instructions_input ] example_outputs_list_comps = [individual_lines_zip_output, merged_dialogue_mp3_output, status_output] example_process_fn_actual = partial(handle_script_processing, OPENAI_API_KEY, async_openai_client, NSFW_API_URL_TEMPLATE) if OPENAI_API_KEY else None # Define the example data directly examples_data = [ [ """[Alice] Hello Bob, this is a test using the detailed configuration method. [Bob] Hi Alice! I'm Bob, and I'll have my own voice settings. [Alice] Let's see how this sounds.""", "tts-1-hd", 300, "Random per Speaker", DEFAULT_GLOBAL_VOICE, 1.0, "" ], [ """[Narrator] Once upon a time, there was a gentle breeze over the hills. [Narrator] The village below prepared for the annual festival as the sun set.""", "gpt-4o-mini-tts", 200, "Detailed Configuration (Per Speaker UI)", DEFAULT_GLOBAL_VOICE, 1.0, "Speak with a gentle, storytelling tone." ], [ """[Solo] This is a quick single‑voice demo for testing purposes.""", "tts-1", 0, "Single Voice (Global)", "fable", 1.2, "" ], ] # Validate example data length against input components length num_inputs_expected = len(example_inputs_list_comps) valid_examples_data_inline = [] for ex_data in examples_data: if len(ex_data) == num_inputs_expected: valid_examples_data_inline.append(ex_data) else: print(f"Warning (Inline Examples): Example data mismatch. Expected {num_inputs_expected}, got {len(ex_data)}. Skipping.") # Directly instantiate gr.Examples if valid data exists if valid_examples_data_inline: if example_process_fn_actual: gr.Examples( examples=valid_examples_data_inline, inputs=example_inputs_list_comps, outputs=example_outputs_list_comps, fn=example_process_fn_actual, cache_examples=False, examples_per_page=5, label="Example Scripts (Click to Load)", # Label is optional if header exists run_on_click=False ) else: gr.Examples( examples=valid_examples_data_inline, inputs=example_inputs_list_comps, examples_per_page=5, label="Example Scripts (Click to Load Inputs)", # Label is optional if header exists ) else: gr.Markdown("

No valid examples could be loaded due to configuration mismatch.

") # --- Launch --- if __name__ == "__main__": if os.name == 'nt': asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy()) demo.queue().launch(debug=True, share=False)