Pixelle-Video is an AI-powered fully automated short video generation engine written in Python that streamlines the entire video creation process from concept to final output. The system is designed to eliminate technical barriers by allowing users to input only a topic, after which the platform automatically handles scriptwriting, AI image and video generation, voice synthesis, background music selection, and final video composition. The project is maintained primarily by contributors puke3615, zhangpeicheng8788-ux, and qingjiuyyds, with puke3615 showing the most activity at 55 tracked events. The repository has grown modestly with one additional fork since July 2026, bringing the total to 3462 forks, and maintains a median issue and pull request response latency of 20.8 hours across 113 tracked items.
The platform features a modular architecture that supports multiple AI models and services. Users can leverage language models including GPT, Qwen, DeepSeek, and Ollama for scriptwriting, while image and video generation can be powered through direct API connections to services like DashScope, OpenAI, Seedream, Seedance, and Kling, or through ComfyUI and RunningHub workflows. Text-to-speech capabilities include support for Edge-TTS and Index-TTS with multi-language voice options. The system supports flexible video dimensions for both vertical and horizontal formats, with multiple visual templates available to customize the final output style.
Recent updates demonstrate active development across several functional areas. The platform added direct API media model configuration in June 2026, enabling users to configure image and video model providers, base URLs, and proxy settings through the web interface. January 2026 brought motion transfer capabilities allowing users to upload reference videos and images for motion migration, alongside new digital human narration and image-to-video pipelines with expanded multilingual TTS voice support. Earlier updates in December 2025 introduced custom material uploads for user-provided photos and videos with AI-powered script generation, Windows integrated package distribution, RunningHub service optimization with parallel processing support, and history tracking with batch video task creation.
The web interface provides an accessible entry point for users without video editing experience. Windows users can utilize the one-click integrated package that includes all dependencies without requiring separate Python, uv, or ffmpeg installations. macOS and Linux users can install from source using the uv package manager and ffmpeg. The video generation workflow follows a clear pipeline of script generation, image planning, frame-by-frame processing, and video synthesis, with each stage supporting customization through different AI models, audio engines, and visual styles.
The repository includes comprehensive documentation at its homepage and video tutorials demonstrating various use cases. Example videos showcase different content categories including humanistic documentary, cultural deconstruction, scientific analysis, personal growth, historical narratives, and knowledge popularization, all generated through single-topic inputs without manual editing. The platform supports specialized modules for digital human narration in multiple languages, image-to-video conversion with cartoon and other styles, and motion transfer effects. The modular design allows flexible combination of atomic capabilities, enabling users to swap image generation, video generation, text-to-speech, and vision-language model components according to their specific needs.
