Pixelle-Video automates short video creation, allowing users to generate complete videos from a simple theme without editing experience. It significantly lowers the barrier to content production by handling scriptwriting, visuals, voiceovers, and assembly.

How It Works

Built on a modular ComfyUI architecture, Pixelle-Video integrates diverse AI capabilities. It generates scripts via LLMs (GPT, Tongyi Qianwen), creates visuals using AI image/video models (ComfyUI/RunningHub), synthesizes voiceovers (Edge-TTS, Index-TTS), and adds background music before final video assembly. This modularity enables flexible model swapping and workflow customization for tailored outputs.

Quick Start & Requirements

For Windows users, a recommended one-click integration package simplifies setup by including all dependencies (Python, uv, ffmpeg) and launching a web interface via start.bat. Source installation for macOS/Linux requires Python, uv, and ffmpeg, followed by cloning the repository and running streamlit run web/app.py. Initial setup involves configuring LLM and image generation API keys in the web UI. Links to uv and ffmpeg installation guides are provided.

Highlighted Details

• Fully automated video generation from a single theme.
• AI-driven scriptwriting, image/video generation, and voiceover synthesis.
• Modular ComfyUI design for flexible AI integration and customization.
• Broad support for multiple LLMs and TTS solutions.
• Customizable visual styles, templates, and audio options.

Maintenance & Community

Community engagement occurs via WeChat and Discord. Users can report issues via GitHub Issues and suggest features through Feature Requests, with project support encouraged via starring.

Licensing & Compatibility

Released under the Apache 2.0 license, permitting commercial use. Users must comply with third-party API terms of service for cloud services.

Limitations & Caveats

Functionality depends on external API keys and may incur costs for cloud AI services. Advanced customization requires ComfyUI familiarity. Local AI model execution is resource-intensive, often needing a GPU.