_    _ _____ __  __ _____ ____      _    _
    / \  | |_   _|  \/  | ____|  _ \    / \  | |
   / _ \ | | | | | |\/| |  _| | |_) |  / _ \ | |
  / ___ \| | | | | |  | | |___|  _ <  / ___ \| |___
 /_/   \_\_| | |_|_|_|  |_|_____|_| \_/_/   \_\_____|

✨ Fitur v2

🤖 AI Engine — Cara Kerja

AITMERAL v2 hadir dengan dua AI engine yang bisa dipilih:

1. MMagic (Primary) — open-mmlab/mmagic

• Repo: https://github.com/open-mmlab/mmagic
• Arsitektur: Berbasis PyTorch + MMEngine + MMCV
• Model yang didukung:
  • RealESRGAN — General purpose super-resolution (x2, x4)
  • SwinIR — Transformer-based, detail recovery superior
  • BasicVSR — Video super-resolution dengan temporal consistency
  • RealBasicVSR — Real-world video SR, handle compression artifacts
• Cara kerja:
  1. Load model checkpoint (otomatis download saat pertama kali)
  2. Preprocess: BGR→RGB, normalize, tensor conversion
  3. Inference: Model forward pass (FP16 default, FP32 optional)
  4. Postprocess: Clamp, denormalize, RGB→BGR, save
  5. Video: Frame-by-frame dengan optional temporal consistency

2. Real-ESRGAN (Alternative) — xinntao/Real-ESRGAN

• Repo: https://github.com/xinntao/Real-ESRGAN
• Arsitektur: RRDBNet + BasicSR framework
• Model yang didukung:
  • realesrgan-x4plus — General purpose 4x
  • realesrgan-x2plus — General purpose 2x
  • realesrgan-anime-x4 — Optimized untuk anime/cartoon
  • realesrgan-video-x4 — Temporal consistency untuk video
  • GFPGAN — Face restoration (opsional)
• Cara kerja:
  1. Load RRDBNet weights
  2. Tiling untuk image besar (memory efficient)
  3. Enhance dengan RealESRGANer class
  4. Support face enhancement via GFPGAN

⚡ Perbandingan

🚀 Instalasi

Metode 1: npm (Paling Mudah — Auto-install Python + Dependencies)

npm install -g aitmeral-enhancer
# Atau one-line via curl:
curl -fsSL https://raw.githubusercontent.com/ZetaGo-Aurum/Aitmeral/main/npm-wrapper/install-npm.sh | bash

Metode 2: Universal Installer (Coba npm → pip → source)

curl -fsSL https://raw.githubusercontent.com/ZetaGo-Aurum/Aitmeral/main/install-universal.sh | bash

Metode 3: Pip (Python Native) — Direkomendasikan untuk AI

# Full AI (both engines)
pip install "aitmeral-enhancer[mmagic,realesrgan]"

# Primary only (MMagic)
pip install "aitmeral-enhancer[mmagic]"

# Alternative only (Real-ESRGAN)
pip install "aitmeral-enhancer[realesrgan]"

# Dengan fitur YouTube:
pip install "aitmeral-enhancer[full]"

Metode 4: Dari Source (Development)

git clone https://github.com/ZetaGo-Aurum/Aitmeral.git
cd Aitmeral
pip install -e ".[mmagic,realesrgan]"
# Atau jalankan helper:
bash install.sh

1. Siapkan dependensi sistem (hanya untuk metode pip/source)

# Linux (Debian/Ubuntu)
sudo apt install ffmpeg python3-pip python3-venv
pip install yt-dlp

# Linux (Fedora)                # Linux (Arch)
sudo dnf install ffmpeg python-pip    sudo pacman -S ffmpeg python-pip
pip install yt-dlp                   pip install yt-dlp

# macOS (Homebrew)
brew install ffmpeg yt-dlp python

# Windows (winget)
winget install Gyan.FFmpeg yt-dlp.yt-dlp Python.Python.3.12

# Termux (Android)
pkg update && pkg install ffmpeg python git
pip install yt-dlp

2. Pasang AI Engine (Setelah install base)

# MMagic (Primary) - butuh MMCV compile
pip install "aitmeral-enhancer[mmagic]"
# Atau manual:
pip install mmcv>=2.0.1 mmengine>=0.10
# mmagic akan auto-install saat pertama dipakai

# Real-ESRGAN (Alternative) - lebih ringan
pip install "aitmeral-enhancer[realesrgan]"
# Atau manual:
pip install basicsr facexlib gfpgan

Catatan: MMagic butuh mmcv yang compile C++ — pastikan punya build tools (gcc, cmake). Real-ESRGAN lebih ringan, pure Python.

🖥️ Cara Pakai

TUI interaktif (default)

aitmeral            # langsung membuka TUI
aitmeral tui        # sama dengan di atas

Panel TUI: Dashboard · File Manager · Enhance & Convert · URL Grabber · Jobs/Queue · System Info.

Web UI lokal

aitmeral server                       # buka http://127.0.0.1:8765
aitmeral server --port 9000 --host 0.0.0.0
aitmeral server --output-dir /mnt/hdd/renders --no-open

Web UI v2: AI Engine selector, Model picker, Process mode (video/image/auto).

CLI — Traditional (ffmpeg)

aitmeral --help
aitmeral -h

# File lokal: 480p → 1080p, output RAW (FFV1 lossless)
aitmeral convert video480.mp4 -S 1080p -p raw --effects denoise:2,sharpen:0.5

# File lokal: hasil kompatibel semua device
aitmeral convert video.mp4 -S 1080p -p superhd --effects all

# Dari YouTube → lossless
aitmeral url "https://www.youtube.com/watch?v=XXXX" -S 1080p -p lossless-x264
aitmeral url "https://www.youtube.com/embed/XXXX"   -S 2160p -p hdr10 --hdr

# SDR → HDR 4K
aitmeral convert video.mp4 -S 2160p -p hdr10 --hdr --effects sharpen:0.5,denoise:1

# Inspeksi media / URL
aitmeral info video.mp4
aitmeral info "https://youtu.be/XXXX"

# Dry-run
aitmeral convert video.mp4 -S 1080p -p raw --dry-run

aitmeral presets      # daftar preset output
aitmeral effects      # daftar efek & shader
aitmeral doctor       # cek sistem

CLI — AI Enhanced (v2 Baru)

# 🖼️ AI Image Upscale 4x (MMagic Primary)
aitmeral convert photo.jpg -p ai-mmagic-realesrgan-x4

# 🖼️ AI Image Upscale 4x → PNG lossless
aitmeral convert photo.jpg -p ai-img-mmagic-x4

# 🎬 AI Video Upscale 4x dengan temporal consistency (MMagic Primary)
aitmeral convert video.mp4 -p ai-mmagic-basicvsr-x4

# 🎬 AI Video Upscale 4x — RealBasicVSR (real-world video)
aitmeral convert video.mp4 -p ai-mmagic-realbasicvsr-x4

# 🎨 AI Video Anime/Cartoon (Real-ESRGAN Alternative)
aitmeral convert anime.mp4 --ai-engine realesrgan --ai-model realesrgan-anime-x4

# 🎭 AI Video General (Real-ESRGAN Alternative)
aitmeral convert video.mp4 --ai-engine realesrgan --ai-model realesrgan-video-x4

# ⚙️ Custom AI settings
aitmeral convert video.mp4 -p ai-mmagic-swinir-x4 \
    --ai-engine mmagic --ai-model swinir-x4 --ai-scale 4 \
    --ai-tile 512 --ai-fp32 --ai-gpu 0

# 🔄 Auto mode (detect image/video)
aitmeral convert input.jpg --mode auto -p ai-mmagic-realesrgan-x4
aitmeral convert input.mp4 --mode auto -p ai-mmagic-basicvsr-x4

Flag AI Baru (aitmeral convert / url)

🎚 Preset Output (aitmeral presets)

Traditional (ffmpeg)

AI-Powered (v2)

🎛 Efek & Shader (aitmeral effects)

📡 Catatan HDR

• SDR→HDR konversi ke cahaya linear → BT.2020 → PQ (SMPTE 2084) lewat zimg, lalu metadata HDR10 pada HEVC. Terbaik di display HDR.
• Sumber HDR dideteksi otomatis: preset RAW/lossless menyimpan HDR 10-bit; preset 8-bit otomatis tone-map ke SDR BT.709.
• Build ffmpeg tanpa zscale didukung — fallback ke 10-bit + signaling metadata (cek aitmeral doctor).

⚠️ Catatan Jujur soal "RAW" & AI

• Preset raw/y4m/lossless-* menjamin tidak ada degradasi dari hasil proses — cocok arsip, re-encode, editing.
• AI upscale tidak menciptakan detail yang tidak ada — hanya recover/enhance detail yang ada dengan neural network.
• File RAW besar (10 menit 1080p = 20–80 GB Y4M). Untuk distribusi gunakan superhd atau hq-hevc.
• AI processing butuh GPU (CUDA) untuk kecepatan optimal. CPU fallback tersedia tapi lambat.

🧰 Troubleshooting

🏗 Struktur Proyek v2

aitmeral/
├── aitmeral/
│   ├── cli.py               # CLI (argparse) — aitmeral --help
│   ├── core/
│   │   ├── ai_engine.py     # 🤖 AI Engine abstraction (MMagic, Real-ESRGAN)
│   │   ├── options.py       # Settings — kontrak CLI/TUI/Web + AI config
│   │   ├── presets.py       # Preset tradisional + AI presets
│   │   ├── filters.py       # Filter-graph tradisional (upscale/shader/HDR)
│   │   ├── encoder.py       # ffmpeg command + hardware accel (NVENC/QSV/AMF)
│   │   ├── downloader.py    # yt-dlp (link embed/shorts dinormalisasi)
│   │   ├── media.py         # ffprobe wrapper
│   │   ├── pipeline.py      # Orkeštrai: tradisional + AI image/video jobs
│   │   ├── jobs.py          # Job engine (TUI & Web)
│   │   └── sysinfo.py       # Deteksi platform/RAM/GPU/encoder/filter
│   ├── tui/app.py           # TUI Textual (file manager, queue live)
│   └── web/
│       ├── server.py        # Web server stdlib + AI API endpoints
│       └── static/index.html# Web UI v2 (AI engine selector, model picker)
├── docs/                    # screenshot TUI (SVG)
├── install.sh               # Installer tradisional
├── install-universal.sh     # Universal installer
├── npm-wrapper/             # npm package wrapper
├── pyproject.toml           # Package config + AI optional deps
└── README.md

⚖️ Lisensi & Etika

MIT License. Gunakan hanya untuk konten yang Anda miliki haknya — mengunduh materi berhak cipta dari YouTube/situs lain tanpa izin dapat melanggar ketentuan layanan mereka dan hukum yang berlaku.