🎬 V-Content Creator

AI-powered viral video content factory — Generate narrated story videos with AI text, AI voice, AI images, and automatic editing. One command, full video.

💸 The "Zero-Cost" Philosophy

This tool was built to battle the paywalls of modern AI content creation. While it supports premium APIs, the recommended workflow is 100% free and doesn't even require an expensive GPU!

By using the Gemini Web capabilities, the script uses Playwright to open a headless browser, log into your Google account (only needed once), and generate both the Stories and Images directly through the Gemini web chat.

No API tokens, no per-generation costs, no RTX 4090 needed. You can run this on a Raspberry Pi if you want. We will keep improving this to democratize content creation.

✨ What it does

Idea → AI Story → AI Voice → AI Images → Subtitles → Final Video

V-Content Creator automates the entire video production pipeline:

1. 📝 Story Generation — AI writes viral stories with hooks, escalation, and twists (100+ Models via LiteLLM: Gemini, OpenAI, DeepSeek, Claude, etc.)
2. 🎙️ Voice Narration — Natural TTS voice (Gemini TTS or ElevenLabs). Includes a free 100% fallback to Edge TTS (no API key needed!)
3. 🖼️ Scene Images — AI-generated images synced to the narration (SDXL local GPU or Gemini Web)
4. 📝 Subtitles — Automatic transcription with faster-whisper
5. 🎬 Video Assembly — FFmpeg with zoompan effects, per-image timing from AI timestamps
6. 📊 YouTube Metadata — Auto-generated title, description, tags, and SEO

🎯 Content Niches

Built-in viral content templates:

Niche	Style
🔍 Misterio Real	True Crime, investigative
🤫 Confesiones	Intimate, confessional
😱 Suspenso Cotidiano	Everyday situations gone wrong
🤖 Ciencia Ficción	Black Mirror-style sci-fi
💔 Drama Humano	Emotional storytelling
🧠 Terror Psicológico	Psychological horror
🌿 Folklore Latam	Latin American legends
⚖️ Venganza	Revenge / poetic justice
🏔️ Supervivencia	Survival stories
💻 Misterio Digital	Internet creepy

Dulus Integration (Highly Recommended)

V-Content Creator is fully compatible as a first-class plugin for Dulus. By running it as a plugin, your AI coding assistant gains full autonomous capabilities to brainstorm ideas, prompt you for formatting constraints (Topic, Format, Duration), and execute the video factory seamlessly.

Installing the Plugin

Open your Dulus REPL and run:

/plugin install video_factory@https://github.com/KevRojo/v-content-creator

Once installed, simply ask your assistant:

"Use the video factory to help me create a short horror story for TikTok."

🚀 Quick Start (Standalone CLI)

1. Clone & Install

git clone https://github.com/YOUR_USERNAME/v-content-creator.git
cd v-content-creator
pip install -r requirements.txt

2. Configure API Keys

cp .env.example .env
# Edit .env with your API keys

Required: At least one of:

• GEMINI_API_KEY — Get it here (free tier available)
• MOONSHOT_API_KEY — Get it here

Optional:

• ELEVENLABS_API_KEY — For premium voice quality
• CHANNEL_NAME — Your channel name for metadata

3. Install FFmpeg

FFmpeg is required for video assembly:

• Windows: winget install ffmpeg or download from ffmpeg.org
• Linux: sudo apt install ffmpeg
• Mac: brew install ffmpeg

4. Generate a Video

# Basic — auto niche, auto duration
python vcontent_creator.py

# Specify a niche and duration
python vcontent_creator.py --niche terror_psicologico --duration 3

# Short format (9:16 vertical, 60s)
python vcontent_creator.py --short

# Use ElevenLabs voice
python vcontent_creator.py --eleven

# Full control
python vcontent_creator.py --count 3 --niche venganza --duration 5 --voice Fenrir --quality high

5. GUI Mode (optional)

pip install PyQt5
python gui.py

🖼️ Image Generation Options

Option A: SDXL Local (GPU required)

Requires an NVIDIA GPU with ~8GB VRAM.

pip install torch diffusers transformers accelerate

Download an SDXL model and place it in:

stable-diffusion-webui-master/models/Stable-diffusion/

Recommended models (download one):

Model	Type	Link
RealVisXL V5.0 Baked VAE	Photorealistic (best)	CivitAI
RealVisXL V5.0 Lightning	Fast photorealistic	CivitAI
Juggernaut XL v11	Multi-purpose premium	CivitAI
SDXL Base 1.0	Official base	HuggingFace

Option B: Gemini Web (No GPU needed, 100% Free!)

This is the recommended, zero-cost method. It uses browser automation via Playwright to generate images directly in the Gemini Web interface (using Imagen 3 under the hood).

• You only need to log in to your Google Account the first time. The script saves your session state.
• Generates high-quality images without needing any local GPU horsepower.

pip install playwright
playwright install chromium
python vcontent_creator.py --gemini-images

📤 Auto-Uploader & Metadata

Every time a video is generated, V-Content Creator builds a comprehensive Metadata JSON file alongside the video (e.g. video_info.json). This file contains the title, optimized description, SEO tags, privacy settings, and even tracking of which platforms it has been uploaded to.

{
  "youtube_title": "Lavanda #Shorts",
  "description": "🎧 \"No abro la puerta de la calle desde el lunes pasado...\"",
  "tags": ["narración", "terror psicológico", "youtube shorts"]
}

Included in the repo is uploader.py, an interactive CLI tool that reads these metadata files and uses Playwright to automatically upload your videos to YouTube Studio and TikTok.

# Launch the interactive uploader menu
python uploader.py

⚙️ CLI Reference

Flag	Description
--count N	Number of stories to generate (default: 1)
--niche NAME	Content niche (use --list-niches to see all)
--duration N	Target duration in minutes
--voice NAME	TTS voice: Charon, Fenrir, Kore, Orus, etc.
--quality LEVEL	Video quality: high, medium, low, minimal
--short	Short format (9:16 vertical, ≤60s)
--model ENGINE	Text model via LiteLLM (e.g. gemini/gemini-2.5-flash, openai/gpt-4o, deepseek/deepseek-chat)
--eleven	Use ElevenLabs TTS instead of Gemini
--gemini-images	Use Gemini Web for image generation
--gemini-web-story	Use Gemini Web for story generation
--list-niches	Show all available niches
--context TEXT	Creative direction for the story

📁 Project Structure

v-content-creator/
├── vcontent_creator.py    # Main engine (text → audio → images → video)
├── uploader.py            # Auto-uploader for YouTube & TikTok (Playwright)
├── gui.py                 # PyQt5 GUI (optional)
├── .env                   # Your API keys (not tracked by git)
├── .env.example           # Template for API keys
├── requirements.txt       # Python dependencies
├── sounds/                # SFX audio files (rain, steps, door)
├── stable-diffusion-webui-master/
│   └── models/
│       └── Stable-diffusion/  # Place SDXL .safetensors models here
├── output/                # Generated videos (auto-created)
└── temp/                  # Working files (auto-created)

🔧 How It Works

1. Story Generation — The AI receives a detailed prompt with the selected niche's tone, hooks, clichés to avoid, and format rules. It generates the story AND image prompts with timestamps.

2. Audio — The story text is split into chunks, synthesized with TTS, and crossfaded into a seamless narration.

3. Images — Each image prompt is generated with a specific timestamp (e.g., IMG1 0:00, IMG2 0:12). Images are resized to match the video format.

4. Video Assembly — Each image becomes a clip with its own zoompan animation matching its specific duration from the AI timestamps. Clips are concatenated with the audio and burned-in subtitles.

📄 License

MIT License — use it, modify it, share it.

🤝 Contributing

Pull requests welcome! Areas that could use help:

• More content niches
• Support for more TTS engines
• Multi-language support
• Thumbnail generation
• Auto-upload to YouTube/TikTok

Donations:

Bitcoin: 1JzatQDn9fMLnKTd3KYgztsLHC95bJEzSN