Pixelle Video

Overview

Pixelle Video compresses the short-form video production pipeline into a single automated workflow. The human bottleneck in short video content — writing scripts, recording voiceover, finding footage, adding subtitles, selecting music, and assembling the final cut — is handled entirely by the system given a topic or prompt.

This is Alibaba AIDC’s entry into the open-source video generation space, released as a direct alternative to closed commercial platforms. The codebase is Python-based and designed for deployment on consumer GPUs (NVIDIA) as well as Alibaba Cloud instances.

What It Produces

A complete short-form video from a text prompt:

1. Script generation — converts the topic into a structured script with scene-by-scene content
2. Voiceover synthesis — text-to-speech with natural prosody, adjustable voice profiles
3. Footage assembly — matches script content to video clips from a built-in or custom footage library
4. Subtitle overlay — synchronized captions with customizable style and positioning
5. Music track — selects from a royalty-free library based on content tone
6. Export — MP4 output in the specified aspect ratio

Use Cases

Content operations automation: Generate draft videos for editorial review faster than any manual pipeline. Edit the script before generation to control the output narrative.

Multilingual content: The TTS pipeline supports multiple languages, enabling the same topic to produce parallel videos in different markets.

Prototype storytelling: Quickly visualize a video concept before committing production resources.

Getting Started

git clone https://github.com/AIDC-AI/Pixelle-Video
cd Pixelle-Video
pip install -r requirements.txt

# Generate a video
python generate.py --topic "5 AI tools that changed how developers work in 2026" --duration 60 --aspect 9:16

Requires Python 3.10+, CUDA-compatible GPU (recommended), and FFmpeg.

Who It’s For

Content teams that produce high-volume short-form video and need to automate the production pipeline. Developers exploring open-source video generation. Teams evaluating alternatives to closed-source video AI platforms before committing to a vendor.