About this project

Pixelle-Video is an open-source AI-powered fully automated short video generation engine. Users simply input a topic, and the system automatically completes video script writing, AI image or video generation, voiceover synthesis, background music addition, and final video composition—all without requiring any video editing experience. Key features include: - AI intelligent script generation, supporting various large language models such as GPT, Qwen, DeepSeek, and Ollama. - AI image and video generation, supporting workflows via ComfyUI and RunningHub, or direct API connections to model providers like DashScope, OpenAI, Seedream, Seedance, and Kling. - Voice synthesis, supporting solutions like Edge-TTS and Index-TTS, along with voice cloning capabilities. - Background music integration, multiple visual style templates, and various video dimensions including portrait and landscape formats. - Extension modules supporting digital human live streaming, image-to-video generation, and motion transfer. - Modular design allowing flexible replacement and combination of each stage (image, video, TTS, VLM). The project provides a Windows one-click integrated package or can be installed from source code, requiring Python package manager uv and ffmpeg configuration. It operates through a Streamlit web interface, where users fill in their LLM API keys and image/video generation service addresses to begin usage. A completely free operation mode is supported (Ollama local LLM + ComfyUI local deployment). The project is licensed under Apache 2.0.