Hi everyone,
We’ve been working in the AI content space for a while, and like many startup teams, we ran into the same problem:creating short marketing videos meant stitching together scripts, prompts, image generation, and voiceovers—often through expensive, credit-based SaaS tools.
Most existing solutions were either costly or too opinionated, making it hard to control style, characters, or branding.
So we built an internal solution to better organize this workflow, and eventually decided to open-source it.
That project is Pixelle-Video .
Pixelle-Video is a node-based AI short-video creation engine for startups and creators who want more control over their video pipeline—without being locked into a single SaaS or cloud provider.
It’s not a black-box, one-click generator, but an engine that lets you combine automation with manual inputs where needed.
What it enables:
- Structured short-video creation starting from a topic or outline, optionally connected to LLMs (DeepSeek, Qwen, OpenAI, etc.)
- Visual and brand control via ComfyUI, using either local workflows or cloud-based services
- Flexible cost and privacy choices , depending on what runs locally vs externally
- Modular components , allowing you to swap LLMs, TTS engines, or image generation backends
The project is Python-based (Streamlit UI) and released under the Apache 2.0 license.
Repo: 
https://github.com/AIDC-AI/Pixelle-Video
We’d love feedback—especially on clarity, setup experience, and whether this kind of engine fits real startup content workflows.
Thanks for supporting open source,
The Pixelle.AI Team