app.py

import tempfile
import edge_tts
import gradio as gr
import asyncio

# Updated language dictionary with clear voice labeling
language_dict = {
    "Amharic": {
        "Ameha": "am-ET-AmehaNeural",
        "Mekdes": "am-ET-MekdesNeural"
    },
    "English": {
        "Ryan": "en-GB-RyanNeural",
        "Clara": "en-CA-ClaraNeural"
    },
    "Tigrinya": {
        "Ameha (Amharic Voice)": "am-ET-AmehaNeural",
        "Mekdes (Amharic Voice)": "am-ET-MekdesNeural"
    }
}

async def text_to_speech_edge(text, language, speaker):
    # Clean speaker name if it's a Tigrinya selection
    if language == "Tigrinya":
        speaker = speaker.replace(" (Amharic Voice)", "")
    
    try:
        voice = language_dict[language][speaker]
    except KeyError:
        error_msg = (f"ስህተት: ድምፅ '{speaker}' ለቋንቋ '{language}' አልተገኘም።"
                    if language in ["Amharic", "Tigrinya"]
                    else f"Error: Voice '{speaker}' not available for '{language}'")
        raise gr.Error(error_msg)

    try:
        communicate = edge_tts.Communicate(text, voice)
        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_file:
            tmp_path = tmp_file.name
            await asyncio.wait_for(communicate.save(tmp_path), timeout=30)
        return tmp_path
        
    except asyncio.TimeoutError:
        error_msg = ("ስህተት: ጊዜ አልቋል። እባክዎ እንደገና ይሞክሩ።"
                    if language in ["Amharic", "Tigrinya"]
                    else "Error: Timeout. Please try again.")
        raise gr.Error(error_msg)
    except Exception as e:
        error_msg = (f"ስህተት: {str(e)}"
                    if language in ["Amharic", "Tigrinya"]
                    else f"Error: {str(e)}")
        raise gr.Error(error_msg)

def update_speakers(language):
    speakers = list(language_dict[language].keys())
    return gr.Dropdown(
        choices=speakers,
        value=speakers[0],
        label=f"Select Speaker {'(Amharic Voices)' if language == 'Tigrinya' else ''}"
    )

with gr.Blocks(title="Amharic, English & Tigrinya TTS", theme=gr.themes.Soft()) as demo:
    gr.HTML("""
    <style>
        h1 { 
            color: #2E86C1; 
            text-align: center;
            background: linear-gradient(45deg, #FF007F, #2E86C1);
            -webkit-background-clip: text;
            -webkit-text-fill-color: transparent;
            margin-bottom: 20px;
        }
        .notice {
            font-size: 0.9em;
            color: #666;
            text-align: center;
            margin: 10px 0;
            font-style: italic;
        }
        .gradio-button {
            background: linear-gradient(45deg, #FF007F, #2E86C1) !important;
            color: white !important;
        }
    </style>
    <center>
        <h1>Amharic, English & Tigrinya Text-to-Speech</h1>
        <div class="notice">
            Note: Tigrinya uses Amharic-accented voices until dedicated models become available
        </div>
    </center>
    """)

    with gr.Row():
        with gr.Column():
            language = gr.Dropdown(
                choices=["Amharic", "English", "Tigrinya"],
                value="Amharic",
                label="Select Language / ቋንቋ ይምረጡ"
            )
            input_text = gr.Textbox(
                lines=5,
                label="Enter Text / ጽሑፍ ያስገቡ",
                placeholder="Type your text here... / ጽሑፍዎን ይጻፉ..."
            )
            speaker = gr.Dropdown(
                label="Select Speaker / አርቲስት ይምረጡ",
                interactive=True
            )
            run_btn = gr.Button(
                value="Generate Audio / ድምፅ ፍጠር",
                variant="primary"
            )

        with gr.Column():
            output_audio = gr.Audio(
                type="filepath",
                label="Generated Audio / የተፈጠረ ድምፅ",
                interactive=False
            )

    # Initialize speakers dropdown
    demo.load(
        fn=lambda: gr.Dropdown(choices=list(language_dict["Amharic"].keys()),
        outputs=speaker
    )

    # Update speakers when language changes
    language.change(
        update_speakers,
        inputs=language,
        outputs=speaker
    )

    run_btn.click(
        text_to_speech_edge,
        inputs=[input_text, language, speaker],
        outputs=output_audio
    )

if __name__ == "__main__":
    demo.launch(server_port=7860, share=False)