import os import json import torch import logging import commons import utils import random import re import gradio as gr from models import SynthesizerTrn from text import text_to_sequence from torch import no_grad, LongTensor from huggingface_hub import hf_hub_download # --- 1. SETUP & CONFIG --- logging.getLogger('numba').setLevel(logging.WARNING) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") os.makedirs("pretrained_models", exist_ok=True) os.makedirs("PICTURE", exist_ok=True) try: hps = utils.get_hparams_from_file(r'config/config.json') except: hps = None # --- 2. ASSETS & DOWNLOADER --- BASE_REPO = "Blue-Archive/VITS-MODELS" def get_category_image(cat_name): file_name = f"{cat_name}.PNG" local_path = os.path.join("PICTURE", file_name) if not os.path.exists(local_path): try: hf_hub_download(repo_id=BASE_REPO, filename=f"PICTURE/{file_name}", local_dir=".") except: return None return local_path def download_assets(char_id, file_name): repo_id = "Blue-Archive/Library-VITS" local_dir = f"pretrained_models/{char_id}" local_path = os.path.join(local_dir, file_name) if not os.path.exists(local_path): try: hf_path = f"VITS-MODELS/pretrained_models/{char_id}/{file_name}" hf_hub_download(repo_id=repo_id, filename=hf_path, local_dir=".") os.makedirs(local_dir, exist_ok=True) if os.path.exists(hf_path): os.replace(hf_path, local_path) except: return None return local_path # --- 3. ENGINE FUNCTIONS --- def tts_execute(char_id, text, lang, speed): if not char_id or not text: return "❌ Pilih karakter dulu!", None model_pth = download_assets(char_id, f"{char_id}.pth") if not model_pth: return "❌ Gagal download model", None net_g = SynthesizerTrn(len(hps.symbols), hps.data.filter_length // 2 + 1, hps.train.segment_size // hps.data.hop_length, n_speakers=hps.data.n_speakers, **hps.model).to(device) utils.load_checkpoint(model_pth, net_g, None) _ = net_g.eval() text = f"[{'JA' if lang == 'Japanese' else 'ZH'}]{text}[{'JA' if lang == 'Japanese' else 'ZH'}]" try: text_norm, _ = text_to_sequence(text, hps.symbols, hps.data.text_cleaners) if hps.data.add_blank: text_norm = commons.intersperse(text_norm, 0) stn_tst = LongTensor(text_norm) with no_grad(): audio = net_g.infer(stn_tst.unsqueeze(0).to(device), LongTensor([stn_tst.size(0)]).to(device), sid=LongTensor([0]).to(device), noise_scale=0.6, noise_scale_w=0.668, length_scale=1.0/speed)[0][0,0].data.cpu().float().numpy() return "✅ Selesai!", (22050, audio) except Exception as e: return f"Error: {e}", None def get_random_jp(): return random.choice(["こんにちは！", "お元気ですか？", "先生、お疲れ様です！", "大好きだよ！", "また明日ね。"]) # --- 4. CSS STYLING (THEME BLUE WITH ORANGE WARNING) --- css = """ /* Loading Gradio Jadi Biru Secara Global */ :root { --primary-600: #1299ff !important; --accent-600: #1299ff !important; } /* Spinner & Progress Bar Biru */ .progress-view, .progress-view > div, [role='progressbar'] > div { background-color: #1299ff !important; } .loading { color: #1299ff !important; } footer {display: none !important;} .gradio-container { background-color: #f8fafc !important; } /* Header & Status */ .ba-header-container { border: 1.5px solid #e1e8f0; border-radius: 12px; padding: 20px; margin-bottom: 12px; background: white; text-align: center; } .ba-header-container h1 { color: #1299ff !important; font-weight: 800 !important; font-size: 32px !important; margin: 0; } .status-container { border: 1.5px solid #e1e8f0; border-radius: 12px; padding: 15px; margin-bottom: 15px; background: white; } /* Dual Scroll Boxes */ .scroll-box { height: 200px !important; overflow-y: auto !important; border: 1px solid #f0f4f8; border-radius: 12px; padding: 10px; background: #fafbfc; margin-bottom: 10px; } .info-scroll-box { height: 200px !important; overflow-y: auto !important; padding: 5px; } /* Character Buttons */ .char-btn { background: white !important; border: 1px solid #e2e8f0 !important; border-left: 5px solid #1299ff !important; text-align: left !important; padding: 10px !important; font-size: 13px !important; margin-bottom: 6px !important; width: 100%; color: #4a5568 !important; border-radius: 8px !important; } /* Picture styling */ .image-display img { object-fit: contain !important; max-height: 150px !important; border-radius: 10px; margin: 0 auto; } /* Warning Card (Warna Kembali ke Orange Semula) */ .warning-card { background: #fff9f0; border: 2px dashed #f5a623; border-radius: 10px; padding: 12px; margin: 10px 0; text-align: center; } /* Character Information Grid */ .info-grid { display: grid !important; grid-template-columns: 1fr 1fr !important; gap: 8px !important; width: 100%; } .info-item { border: 1px solid #f0f4f8; border-radius: 10px; padding: 10px; border-left: 4px solid #1299ff; background: white; box-shadow: 0 2px 4px rgba(0,0,0,0.02); } /* Buttons & Inputs */ .gen-btn { background: #1299ff !important; color: white !important; font-weight: 700 !important; border-radius: 12px !important; height: 48px !important; width: 100%; border: none !important; } .jp-btn { background: #f8fafc !important; border: 1px solid #cbd5e1 !important; font-size: 11px !important; border-radius: 8px !important; margin-bottom: 8px; width: 100%; font-weight: 700 !important; } """ # --- 5. DATA PREP --- try: if not os.path.exists("pretrained_models/info.json"): hf_hub_download(repo_id="Blue-Archive/Library-VITS", filename="VITS-MODELS/pretrained_models/info.json", local_dir=".") os.replace("VITS-MODELS/pretrained_models/info.json", "pretrained_models/info.json") with open("pretrained_models/info.json", "r", encoding="utf-8") as f: MODELS_INFO = json.load(f) except: MODELS_INFO = {} CATEGORIES = {} for m_id, info in MODELS_INFO.items(): cat = info['title'].split("-")[0].strip() if cat == "Overwatch 2": continue if cat not in CATEGORIES: CATEGORIES[cat] = [] CATEGORIES[cat].append((m_id, info)) # --- 6. UI CONSTRUCTION --- with gr.Blocks(css=css, title="VITS MODELS") as demo: with gr.Column(elem_classes="slim-card"): gr.HTML("""

VITS MODELS

Voice Library System

System Status
Engine : READY ✅
System : Online
""") with gr.Tabs(): for idx, cat_name in enumerate(sorted(CATEGORIES.keys())): with gr.Tab(f"Model {idx}"): gr.Markdown(f"### 📂 {cat_name}") cat_img = get_category_image(cat_name) char_preview = gr.Image(value=cat_img, show_label=False, interactive=False, elem_classes="image-display") sel_char_id = gr.State("") char_display = gr.Markdown("📍 *Silakan pilih karakter...*") # Selection Scroll Box with gr.Column(elem_classes="scroll-box"): for m_id, info in CATEGORIES[cat_name]: name_en = info.get('name_en', m_id) btn = gr.Button(f"👤 {name_en}", elem_classes="char-btn") def update_sel(c_id=m_id, n_en=name_en, cov=info.get('cover')): img = download_assets(c_id, cov) return c_id, f"📍 Selected: **{n_en}**", img btn.click(fn=update_sel, outputs=[sel_char_id, char_display, char_preview]) # Warning Card (Warna Kembali ke Orange/Jingga) gr.HTML("""
🔖 PERINGATAN MINNA 🔖
Klik character, ketik teks, lalu Generate! ✨
""") txt_in = gr.TextArea(label="Input Text", placeholder="Masukkan teks...", lines=3) gr.Button("🎲 INPUTS RANDOM TEXT 🎲", elem_classes="jp-btn").click(get_random_jp, outputs=[txt_in]) with gr.Row(): lang_drop = gr.Dropdown(choices=["Japanese", "Chinese"], label="Lang", value="Japanese") spd = gr.Slider(0.5, 2.0, 1.0, step=0.1, label="Speed Audio") btn_gen = gr.Button("🎐 GENERATE VOICE 🎐", elem_classes="gen-btn") aud_out = gr.Audio(label="Voice Output") # Info Accordion with Scrollable Grid with gr.Accordion("📑 Character Information 📑", open=False): with gr.Column(elem_classes="info-scroll-box"): with gr.Column(elem_classes="info-grid"): for m_id, info in CATEGORIES[cat_name]: name_en = info.get('name_en', m_id) gr.HTML(f"""
{name_en}
Character
""") btn_gen.click(fn=tts_execute, inputs=[sel_char_id, txt_in, lang_drop, spd], outputs=[gr.Textbox(visible=False), aud_out]) gr.HTML("
🌥️ CREATED BY MUTSUMI 🌥️
") if __name__ == "__main__": demo.queue().launch()