# Pixelle-Video — AI Automated Short Video Engine > Open-source AI engine for fully automated short video production, from script generation to final render, with ComfyUI and TTS integration. ## Install Save as a script file and run: # Pixelle-Video — AI Automated Short Video Engine ## Quick Use ```bash git clone https://github.com/ATH-MaaS/Pixelle-Video.git cd Pixelle-Video pip install -r requirements.txt python main.py --config config.yaml ``` ## 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 --- Source: https://tokrepo.com/en/workflows/asset-31ba1bd2 Author: Script Depot