### Install FFmpeg using winget

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/README_en.md

Use this command in PowerShell (as Administrator) to install FFmpeg on Windows.

```powershell
winget install ffmpeg
```

--------------------------------

### Verify FFmpeg Installation

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/README_en.md

After installation, run this command in your terminal to confirm FFmpeg is working correctly.

```bash
ffmpeg -version
```

--------------------------------

### Initialize ViralCutter

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/ViralCutter.ipynb

This snippet shows the command to start the ViralCutter application using its virtual environment. It also indicates that UserWarnings can be ignored and provides the local and public URLs for accessing the web UI.

```bash
cd /content/ViralCutter
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python webui/app.py
```

--------------------------------

### Install ViralCutter and Dependencies

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/ViralCutter.ipynb

Installs ViralCutter from GitHub, sets up a virtual environment using UV, and installs all necessary Python libraries. It includes specific version pinning for critical libraries like Transformers and PyTorch to ensure compatibility.

```python
#@title 🛠️ Instalação
import os
import subprocess
import shutil
from IPython.display import clear_output

# 1. Limpeza TOTAL
print("🧹 Limpando instalação anterior...")
%cd /content
if os.path.exists("ViralCutter"):
    shutil.rmtree("ViralCutter")

!git clone -b dev https://github.com/RafaelGodoyEbert/ViralCutter.git
%cd /content/ViralCutter

print("⏳ Instalando gerenciador UV e drivers do sistema...")

# 2. Instalar UV e Drivers Linux
subprocess.run(['pip', 'install', 'uv'], check=True)
subprocess.run('sudo apt update -y && sudo apt install -y libcudnn8 ffmpeg xvfb', shell=True, check=True)

# 3. Criar Ambiente Virtual
print("⏳ Criando ambiente virtual...")
subprocess.run(['uv', 'venv', '.venv'], check=True)

# 4. INSTALAÇÃO DAS DEPENDÊNCIAS
print("⏳ Instalando Bibliotecas...")

# Passo A: WhisperX e Requisitos Básicos (Deixe instalar o que quiserem)
cmds_fase_1 = [
    "uv pip install --python .venv git+https://github.com/m-bain/whisperx.git",
    "uv pip install --python .venv -r requirements.txt",
    "uv pip install --python .venv yt-dlp pytubefix",
    "uv pip install yt_dlp"
]

for cmd in cmds_fase_1:
    subprocess.run(cmd, shell=True, check=True)

# Passo B: CORREÇÃO DE ALINHAMENTO E GEMINI
# - google-generativeai: Para o Gemini funcionar
# - pandas: Para separar palavras
# - transformers==4.46.3: VERSÃO CRÍTICA. Versões mais novas exigem Torch 2.6 e quebram o alinhamento.
# - accelerate: Ajuda no carregamento do modelo
print("🔨 Aplicando downgrade estratégico no Transformers...")
extra_libs = [
    "uv pip install --python .venv google-generativeai",
    "uv pip install --python .venv pandas",
    "uv pip install --python .venv onnxruntime-gpu",
    "uv pip install --python .venv transformers==4.46.3 accelerate>=0.26.0"
]

for cmd in extra_libs:
    subprocess.run(cmd, shell=True, check=True)

# Passo C: O MARTELO FINAL (Torch 2.3.1 Estável)
# Reinstalamos por último para garantir que nada atualizou ele sem querer
print("🔨 Forçando versão estável do Torch (2.3.1)...")
cmd_fix_torch = (
    "uv pip install --python .venv "
    "torch==2.3.1+cu121 torchvision==0.18.1+cu121 torchaudio==2.3.1+cu121 "
    "--index-url https://download.pytorch.org/whl/cu121"
)
subprocess.run(cmd_fix_torch, shell=True, check=True)

# Passo D: Trava do Numpy
print("🔨 Travando Numpy...")
subprocess.run("uv pip install --python .venv 'numpy<2.0' setuptools==69.5.1", shell=True, check=True)

# Passo E: Correção de Visão Computacional (CRÍTICO)
print("🔨 Corrigindo InsightFace e MediaPipe...")

# 1. Garante que o InsightFace tenha o motor necessário (evita que ele falhe)
subprocess.run("uv pip install --python .venv insightface onnxruntime-gpu", shell=True, check=True)

# 2. Limpeza Radical e Reinstalação do MediaPipe
# Remove versões quebradas
subprocess.run("uv pip uninstall --python .venv mediapipe protobuf flatbuffers", shell=True)

# 3. Instala versão FLEXÍVEL respeitando limites críticos
# mediapipe>=0.10.0: Pega a versão mais recente compatível com Py3.12
# protobuf>=3.20,<5.0: Garante compatibilidade sem quebrar o sistema antigo
subprocess.run("uv pip install --python .venv 'mediapipe>=0.10.0' 'protobuf>=3.20,<5.0' 'flatbuffers>=2.0'", shell=True, check=True)

# 5. Configurar Monitor
os.system('Xvfb :1 -screen 0 2560x1440x8 &')
os.environ['DISPLAY'] = ':1.0'

clear_output()
print("✅ Instalação V7 Finalizada!")
print("- Transformers 4.46.3 (Compatível com Alinhamento): INSTALADO")
print("- Torch 2.3.1: ATIVO")

```

--------------------------------

### Run ViralCutter Web UI

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/ViralCutter.ipynb

Configures the display environment and runs the ViralCutter web interface using the installed virtual environment. It forces Matplotlib to use the 'Agg' backend to avoid conflicts with Colab's inline backend.

```python
#@title 🚀 Configuração e Execução
import os

%cd /content/ViralCutter

# 1. Configura o "Display" falso (para evitar erros de interface gráfica)
os.system('Xvfb :1 -screen 0 2560x1440x8 &')
os.environ['DISPLAY'] = ':1.0'

# 2. CORREÇÃO DO ERRO ATUAL:
# Forçamos o Matplotlib a usar o backend 'Agg' (modo sem cabeça/headless)
# Isso ignora o 'module://matplotlib_inline.backend_inline' que o Colab força e quebra o script.
os.environ['MPLBACKEND'] = 'Agg'

print("🚀 Iniciando ViralCutter usando o ambiente virtual...")
print("⚠️ Ignore os avisos de 'UserWarning', aguarde o link public URL.")

# Executa usando o Python do ambiente virtual
!/content/ViralCutter/.venv/bin/python webui/app.py --colab

```

--------------------------------

### Run ViralCutter Web UI

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/README_en.md

Double-click this batch file to launch the ViralCutter web interface in your browser.

```batch
run_webui.bat
```

--------------------------------

### Run ViralCutter CLI

Source: https://github.com/rafaelgodoyebert/viralcutter/blob/main/README_en.md

Use this Python command to run the command-line interface version of ViralCutter.

```python
python main_improved.py
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.