wilson1yan/VideoGPT

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1081 stars · 129 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

VideoGPT is a PyTorch implementation of a video generation model that combines a VQ-VAE with 3D convolutions and axial attention with a GPT-style autoregressive transformer over discrete latents. It serves as a minimalistic, reproducible reference for transformer-based video generation research.

Use cases

• generate videos with a transformer model
• train a VQ-VAE on video data
• reproduce the VideoGPT paper results
• experiment with autoregressive video generation
• learn how VQ-VAE and GPT combine for video synthesis

When to choose

• you need a simple reference implementation of likelihood-based video generation
• you want to train or fine-tune a VQ-VAE plus transformer on your own video dataset
• you are doing research on discrete latent video models

When to avoid

• you need production-ready, actively maintained video generation
• you want large-scale text-to-video generation
• you lack a GPU or cannot manage older CUDA/PyTorch dependencies

Facets

library · maturity maintenance

machine-learning deep-learning video-processing transformers deep-learning computer-vision machine-learning python vq-vae video-generation gpt autoregressive-model research-code video gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="wilson1yan/VideoGPT")