Update app.py

app.py

@@ -12,6 +12,7 @@ import requests
 import base64
 import wave
 
+# EDGE voices
 VOICES_FILE = Path(__file__).parent / "voices.json"
 with open(VOICES_FILE, encoding="utf-8") as f:
     voices_data = json.load(f)
@@ -19,10 +20,10 @@ with open(VOICES_FILE, encoding="utf-8") as f:
 languages = sorted(voices_data.keys())
 def get_genders(lang):
     return sorted(voices_data.get(lang, {}).keys())
+
 def get_edge_voices(lang, gender):
     return [v["display_name"] for v in voices_data.get(lang, {}).get(gender, [])]
 
-# Map Edge display_name → voice_code
 edge_voice_map = {
     v["display_name"]: v["voice_code"]
     for block in voices_data.values()
@@ -30,12 +31,16 @@ edge_voice_map = {
     for v in gender_block
 }
 
-# Google Gemini voice map (ví dụ, bạn nên có 1 file json chứa voice Gemini, hoặc hardcode một số mẫu)
-google_gemini_voices = [
-    "vi-VN-Wavenet-A", "vi-VN-Wavenet-B", "en-US-Wavenet-D", "Puck"
+# Gemini voices (hardcoded)
+GEMINI_VOICES = [
+    "Zephyr", "Puck", "Charon", "Kore", "Fenrir", "Leda", "Orus", "Aoede",
+    "Callirrhoe", "Autonoe", "Enceladus", "Iapetus", "Umbriel", "Algieba",
+    "Despina", "Erinome", "Algenib", "Rasalgethi", "Laomedeia", "Achernar",
+    "Alnilam", "Schedar", "Gacrux", "Pulcherrima", "Achird",
+    "Zubenelgenubi", "Vindemiatrix", "Sadachbia", "Sadaltager", "Sulafat"
 ]
 def get_gemini_voices():
-    return google_gemini_voices
+    return GEMINI_VOICES
 
 MAX_CHARS = 5000  # Edge-TTS giới hạn ~5000 ký tự
 
@@ -120,7 +125,6 @@ def run_gemini_tts(text, model, api_key, voice_name, out_path):
     resp.raise_for_status()
     b64 = resp.json()["candidates"][0]["content"]["parts"][0]["inlineData"]["data"]
     pcm_data = base64.b64decode(b64)
-    # Lưu thành WAV rồi chuyển sang MP3
     wav_path = out_path.replace('.mp3', '.wav')
     with wave.open(wav_path, "wb") as wf:
         wf.setnchannels(1)
@@ -145,7 +149,7 @@ def build_srt(segments, durations, srt_path):
             f.write(f"{i}\n{start} --> {end}\n{seg.strip()}\n\n")
             cum += dur
 
-def generate_tts(text, engine, lang, gender, voice_name, rate, pitch, api_key, gemini_model):
+def generate_tts(text, engine, lang, gender, voice_name, rate, pitch, api_key, gemini_model, gemini_voice):
     if not text.strip():
         return None, None
     segments = split_text(text)
@@ -160,7 +164,7 @@ def generate_tts(text, engine, lang, gender, voice_name, rate, pitch, api_key, g
         elif engine == "Google API":
             if not api_key:
                 raise Exception("Bạn phải nhập API key Google Gemini TTS!")
-            run_gemini_tts(seg, gemini_model, api_key, voice_name, seg_path)
+            run_gemini_tts(seg, gemini_model, api_key, gemini_voice, seg_path)
         seg_files.append(seg_path)
         durations.append(get_audio_duration(seg_path))
     out_audio = "LyTranTTS.mp3"
@@ -169,7 +173,7 @@ def generate_tts(text, engine, lang, gender, voice_name, rate, pitch, api_key, g
     build_srt(segments, durations, srt_file)
     return out_audio, srt_file
 
-def preview_voice(engine, lang, gender, voice_name, rate, pitch, api_key, gemini_model):
+def preview_voice(engine, lang, gender, voice_name, rate, pitch, api_key, gemini_model, gemini_voice):
     text = "Xin chào! Đây là ứng dụng chuyển văn bản thành giọng, nói được phát triển bởi Lý Trần." if "Việt" in lang else "Hello, This is a Text to Speech, Speech App developed by Ly Tran"
     out = "LyTranTTS.mp3"
     if engine == "Edge":
@@ -180,14 +184,18 @@ def preview_voice(engine, lang, gender, voice_name, rate, pitch, api_key, gemini
     elif engine == "Google API":
         if not api_key:
             raise Exception("Bạn phải nhập API key Google Gemini TTS!")
-        run_gemini_tts(text, gemini_model, api_key, voice_name, out)
+        run_gemini_tts(text, gemini_model, api_key, gemini_voice, out)
     return out
 
 def on_engine_change(engine):
-    if engine == "Google API":
-        return gr.update(visible=True), gr.update(visible=True), gr.update(choices=get_gemini_voices(), value=get_gemini_voices()[0]), gr.update(visible=True)
-    else:
-        return gr.update(visible=False), gr.update(visible=False), gr.update(choices=get_edge_voices(DEFAULT_LANG, get_genders(DEFAULT_LANG)[0]), value=get_edge_voices(DEFAULT_LANG, get_genders(DEFAULT_LANG)[0])[0]), gr.update(visible=False)
+    # Show/hide Gemini fields when engine changes
+    show_gemini = engine == "Google API"
+    return (
+        gr.update(visible=show_gemini),   # api_key_input
+        gr.update(visible=show_gemini),   # gemini_model_input
+        gr.update(visible=not show_gemini),   # voice_dd (Edge voice)
+        gr.update(visible=show_gemini),   # gemini_voice_dd
+    )
 
 def on_language_change(lang):
     genders = get_genders(lang)
@@ -207,7 +215,83 @@ def on_gender_change(lang, gender):
 DEFAULT_LANG = "Đa Ngôn Ngữ"
 DEFAULT_MODEL = "gemini-2.5-flash-preview-tts"
 
-custom_css = """ ... """  # giữ nguyên phần CSS như bạn có
+custom_css = """
+:root {
+    --primary: #4f46e5;
+    --secondary: #f9fafb;
+    --accent: #10b981;
+    --text: #1f2937;
+    --border: #e5e7eb;
+}
+body {
+    font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
+}
+.gradio-container {
+    max-width: 900px !important;
+    margin: 0 auto;
+    background-color: white;
+    border-radius: 12px;
+    box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);
+}
+h1 {
+    color: var(--primary) !important;
+    font-weight: 700 !important;
+    margin-bottom: 0.5rem !important;
+}
+.description {
+    color: var(--text) !important;
+    margin-bottom: 1.5rem !important;
+    font-size: 1rem !important;
+}
+.input-section, .output-section {
+    background-color: var(--secondary);
+    padding: 1.5rem;
+    border-radius: 8px;
+    margin-bottom: 1.5rem;
+    border: 1px solid var(--border);
+}
+.input-section label, .output-section label {
+    font-weight: 600 !important;
+    color: var(--text) !important;
+}
+.tab {
+    background-color: white !important;
+    border-radius: 8px !important;
+    padding: 1rem !important;
+}
+button {
+    background-color: var(--primary) !important;
+    color: white !important;
+    border: none !important;
+    border-radius: 6px !important;
+    padding: 0.5rem 1rem !important;
+    font-weight: 500 !important;
+    transition: all 0.2s !important;
+}
+button:hover {
+    background-color: #4338ca !important;
+    transform: translateY(-1px) !important;
+}
+.preview-btn {
+    background-color: var(--accent) !important;
+}
+.preview-btn:hover {
+    background-color: #0d9b6e !important;
+}
+.slider-container {
+    margin-top: 1rem;
+}
+.slider-container label {
+    margin-bottom: 0.5rem;
+    display: block;
+}
+footer {
+    text-align: center;
+    margin-top: 2rem;
+    color: #6b7280;
+    font-size: 0.875rem;
+}
+"""
 
 with gr.Blocks(title="LyTranTTS - Phiên bản miễn phí", css=custom_css) as demo:
     with gr.Row():
@@ -228,7 +312,7 @@ with gr.Blocks(title="LyTranTTS - Phiên bản miễn phí", css=custom_css) as
             ["Edge", "Google", "Google API"], 
             label="Chọn Engine TTS", 
             value="Edge",
-            info="Edge TTS hỗ trợ nhiều giọng đọc hơn, Google TTS đơn giản và nhanh, Google API dùng API key"
+            info="Edge TTS hỗ trợ nhiều giọng đọc hơn, Google TTS đơn giản và nhanh, Google API dùng API key Gemini"
         )
         text_input = gr.Textbox(
             label="Nhập văn bản cần chuyển đổi", 
@@ -330,12 +414,12 @@ with gr.Blocks(title="LyTranTTS - Phiên bản miễn phí", css=custom_css) as
     # Generate
     gen_btn.click(
         fn=generate_tts,
-        inputs=[text_input, engine, lang_dd, gender_dd, voice_dd, rate_sl, pitch_sl, api_key_input, gemini_model_input],
+        inputs=[text_input, engine, lang_dd, gender_dd, voice_dd, rate_sl, pitch_sl, api_key_input, gemini_model_input, gemini_voice_dd],
         outputs=[out_audio, out_srt]
     )
     preview_btn.click(
         fn=preview_voice,
-        inputs=[engine, lang_dd, gender_dd, voice_dd, rate_sl, pitch_sl, api_key_input, gemini_model_input],
+        inputs=[engine, lang_dd, gender_dd, voice_dd, rate_sl, pitch_sl, api_key_input, gemini_model_input, gemini_voice_dd],
         outputs=[prev_audio]
     )