Tutorial Pipeline

A bilingual tutorial video pipeline that automatically processes videos by downloading, transcribing, cleaning transcripts, generating voice-overs, and assembling the final video.

Prerequisites

• OS: Windows 10 or 11
• Privileges: Administrator access (required for the initial setup script to install system dependencies via winget)
• Hardware: An NVIDIA GPU with at least 4GB VRAM is highly recommended for faster Whisper transcription and local TTS generation, though it will fall back to CPU if no GPU is detected.

---

1. Initial Setup

To get started on a fresh machine, open PowerShell as an Administrator and run the setup script. This script will automatically install Python 3.11, FFmpeg, Git, and uv. It will also create a virtual environment, install the required Python packages, and set up the local VieNeu-TTS engine.

.\setup.ps1

2. Configuration

Once the setup is complete, you need to configure your environment variables and voice reference:

1. API Keys: Go to the config folder. The setup script created a .env file based on .env.template. Open the .env file and add your required API keys (e.g., Anthropic Claude or Antigravity).
2. Voice Clone Reference: If you are using local TTS cloning, place a clean audio sample of the voice you want to clone at voice_refs\reference.wav.

---

3. How to Use

You can run the pipeline either through the Command Line or via the Web UI.

Option A: Command Line Interface (CLI)

To run a video through the entire pipeline automatically, use the run.ps1 script. Place your source video in the inbox folder, then run:

.\scripts\run.ps1 -VideoPath inbox\your_video.mp4

(Note: The pipeline also supports fetching videos automatically if you configure it to download from YouTube via download_youtube.py)

Option B: Web UI

If you prefer a graphical interface to manage your tasks, you can use the built-in Web UI.

1. Start the UI server by running:

   .\start_ui.ps1
   

2. Open your web browser and navigate to: http://127.0.0.1:5000

---

Pipeline Overview

1. 📥 Download: Uses download_youtube.py to fetch videos (or grabs local files).
2. 📝 Transcribe: Uses Whisper (transcribe.py) to generate an accurate transcript of the audio.
3. ✨ Clean & Translate: Uses LLMs (clean_script.py) to refine the transcript, fix errors, and translate if needed.
4. 🗣️ Generate Voice: Uses VieNeu-TTS (generate_voice_local.py or cloud alternative) to synthesize a new voice-over based on your reference audio.
5. 🎬 Assemble: Uses FFmpeg (assemble.py) to stitch the new audio back into the original video, outputting the final file to the output folder.