| |
| """Gradio Web Studio / Hugging Face Space для модели Luna 1.0 (RU / EN Chat + Audio). |
| |
| Запуск: |
| pip install gradio |
| python app.py |
| """ |
|
|
| from __future__ import annotations |
|
|
| import os |
| from pathlib import Path |
|
|
| from transformers import AutoModel |
|
|
| MODEL_DIR = os.environ.get("LUNA_MODEL_PATH", str(Path(__file__).resolve().parent)) |
|
|
|
|
| def create_demo(): |
| import gradio as gr |
|
|
| luna = AutoModel.from_pretrained(MODEL_DIR, trust_remote_code=True) |
| voice_names = list(luna.list_voices().keys()) |
|
|
| def reply_and_speak(message, history, voice, language, max_tokens, temperature, add_chime): |
| history = history or [] |
| out = luna.chat_and_speak( |
| prompt=message, |
| history=history, |
| language=language, |
| voice=voice, |
| chime=True if add_chime else None, |
| max_new_tokens=int(max_tokens), |
| temperature=float(temperature), |
| ) |
| new_history = list(history) + [ |
| {"role": "user", "content": message}, |
| {"role": "assistant", "content": out.text}, |
| ] |
| return "", new_history, (out.sample_rate, out.waveform) |
|
|
| def synthesize_only(text, voice, language, speed, add_chime, ambient): |
| amb = None if ambient == "none" else ambient |
| out = luna.speak( |
| text=text, |
| language=language, |
| voice=voice, |
| speaking_rate=float(speed), |
| chime=bool(add_chime), |
| ambient=amb, |
| ) |
| info = f"Длительность / Duration: {out.duration_sec:.2f}s | Язык / Lang: {out.language} | {out.sample_rate} Hz" |
| return (out.sample_rate, out.waveform), info |
|
|
| def create_sfx(effect, duration): |
| out = luna.create_audio_fx(effect=effect, duration_sec=float(duration)) |
| return (out.sample_rate, out.waveform) |
|
|
| with gr.Blocks(title="Luna 1.0 — RU/EN Chat & Audio Studio") as demo: |
| gr.Markdown( |
| "# 🌙 Luna 1.0 — Bilingual (RU / EN) Chat & Neural Audio Model\n" |
| "Двуязычная диалоговая модель (русский + английский) с мгновенной голосовой озвучкой ответов, " |
| "пресетами тембра и генератором атмосферного аудио." |
| ) |
|
|
| with gr.Tab("💬 Диалог + Озвучка (Chat & Speak)"): |
| chatbot = gr.Chatbot(type="messages", height=360, label="Luna 1.0") |
| audio_out = gr.Audio(label="Голосовой ответ Луны / Spoken Reply", autoplay=True) |
| msg = gr.Textbox( |
| label="Ваше сообщение / Your message", |
| placeholder="Напишите по-русски или по-английски… / Type in Russian or English…", |
| ) |
| with gr.Row(): |
| voice_dd = gr.Dropdown(voice_names, value="luna_warm", label="Голос / Voice Preset") |
| lang_dd = gr.Dropdown(["auto", "ru", "en"], value="auto", label="Язык / Language") |
| chime_cb = gr.Checkbox(value=False, label="Приветственный аккорд (Chime)") |
| with gr.Row(): |
| max_tok = gr.Slider(32, 320, value=140, step=16, label="Max new tokens") |
| temp = gr.Slider(0.1, 1.2, value=0.7, step=0.05, label="Temperature") |
| send_btn = gr.Button("Отправить и озвучить / Send & Speak", variant="primary") |
|
|
| send_btn.click( |
| reply_and_speak, |
| inputs=[msg, chatbot, voice_dd, lang_dd, max_tok, temp, chime_cb], |
| outputs=[msg, chatbot, audio_out], |
| ) |
| msg.submit( |
| reply_and_speak, |
| inputs=[msg, chatbot, voice_dd, lang_dd, max_tok, temp, chime_cb], |
| outputs=[msg, chatbot, audio_out], |
| ) |
|
|
| with gr.Tab("🎙️ Озвучка текста RU/EN (TTS Studio)"): |
| tts_text = gr.Textbox( |
| lines=4, |
| value="Привет! Меня зовут Luna 1.0. Hello! I can speak both Russian and English seamlessly!", |
| label="Текст для озвучки / Text to synthesize", |
| ) |
| with gr.Row(): |
| tts_voice = gr.Dropdown(voice_names, value="luna_warm", label="Пресет голоса") |
| tts_lang = gr.Dropdown(["auto", "ru", "en"], value="auto", label="Язык") |
| tts_speed = gr.Slider(0.75, 1.35, value=1.0, step=0.05, label="Скорость речи") |
| with gr.Row(): |
| tts_chime = gr.Checkbox(value=True, label="Добавить аккорд Luna") |
| tts_amb = gr.Dropdown(["none", "lunar", "warm", "focus"], value="lunar", label="Эмбиент-фон") |
| tts_btn = gr.Button("Синтезировать аудио / Generate Speech", variant="primary") |
| tts_audio = gr.Audio(label="Результат WAV") |
| tts_meta = gr.Textbox(label="Параметры сигнала", interactive=False) |
| tts_btn.click( |
| synthesize_only, |
| inputs=[tts_text, tts_voice, tts_lang, tts_speed, tts_chime, tts_amb], |
| outputs=[tts_audio, tts_meta], |
| ) |
|
|
| with gr.Tab("🌌 Звуки и Эмбиент (Audio FX)"): |
| with gr.Row(): |
| sfx_kind = gr.Dropdown( |
| ["chime", "outro", "lunar_pad", "success", "thinking", "scifi_pulse", "heartbeat", "gentle_rain"], |
| value="lunar_pad", |
| label="Тип аудиоэффекта / Sound Structure", |
| ) |
| sfx_dur = gr.Slider(1.0, 10.0, value=3.5, step=0.5, label="Длительность (сек)") |
| sfx_btn = gr.Button("Создать звук / Create Audio", variant="primary") |
| sfx_audio = gr.Audio(label="Сгенерированный звук") |
| sfx_btn.click(create_sfx, inputs=[sfx_kind, sfx_dur], outputs=[sfx_audio]) |
|
|
| return demo |
|
|
|
|
| if __name__ == "__main__": |
| app = create_demo() |
| app.launch(server_name="0.0.0.0", server_port=int(os.environ.get("PORT", "7860"))) |
|
|
