Pixelle-Video — AI Automated Short Video Engine

Introduction

Pixelle-Video is an open-source AI engine that automates the entire short video production pipeline. It takes a topic or script and produces a finished video with voiceover, visuals, subtitles, and background music without manual editing.

What Pixelle-Video Does

• Generates scripts from topics using LLM integration
• Produces AI voiceover via configurable TTS backends
• Creates or selects visuals through ComfyUI image generation
• Adds synchronized subtitles and background music
• Outputs ready-to-publish short videos in standard formats

Architecture Overview

Pixelle-Video is built as a modular pipeline in Python. Each stage (script, audio, visual, subtitle, composition) runs as an independent module orchestrated by a central workflow engine. ComfyUI handles image generation, FFmpeg handles media compositing, and LLM providers handle script creation. Configuration files define the pipeline steps and model choices.

Self-Hosting & Configuration

• Requires Python 3.10+ and FFmpeg installed locally
• Configure LLM provider keys in the YAML config
• Set up ComfyUI locally or point to a remote instance for image generation
• TTS backend is configurable (multiple engines supported)
• Output resolution, format, and duration are adjustable via config

Key Features

• End-to-end automated video production from text input
• Pluggable LLM, TTS, and image generation backends
• ComfyUI integration for high-quality AI-generated visuals
• Batch processing for multiple videos in sequence
• Subtitle generation with timing synchronization

Comparison with Similar Tools

• MoneyPrinterTurbo — focuses on monetization workflows; Pixelle-Video is more modular and extensible
• Revideo — code-driven video editing; Pixelle-Video is fully automated from script to output
• FFmpeg — low-level media tool; Pixelle-Video orchestrates FFmpeg as part of a higher-level pipeline
• OpenMontage — AI production system; Pixelle-Video targets short-form content specifically

FAQ

Q: What LLM providers does it support? A: It supports OpenAI-compatible APIs including local models via Ollama or vLLM.

Q: Do I need a GPU? A: A GPU is recommended for ComfyUI image generation but not strictly required for the pipeline itself.

Q: What video formats are supported? A: MP4 is the default output format; other FFmpeg-supported formats can be configured.

Q: Can I use my own images instead of AI-generated ones? A: Yes, you can provide a folder of images and the pipeline will use them instead of generating new ones.

Sources

• https://github.com/ATH-MaaS/Pixelle-Video
• https://github.com/ATH-MaaS/Pixelle-Video/blob/main/README.md