# openai/Video-Pre-Training

Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos

Repository: https://github.com/openai/Video-Pre-Training
Canonical: https://ross.abutalabs.com/products/video-pre-training
Language: Python
License: MIT
License Family: permissive
Last push: 2025-09-03T21:51:23+00:00

## Health v2 (maintenance only)
Score: 41/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 40, release rhythm 8, longevity 100
- inputs: {"age_days": 1533, "days_push": 364, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1737, forks 171 (observed 2026-08-28T04:05:29.870777+00:00)

## What it is
OpenAI's Video PreTraining (VPT) codebase for learning Minecraft agents by watching unlabeled online videos, including behavioral cloning and reinforcement learning pipelines. It ships pre-trained foundation and fine-tuned agent models runnable in the MineRL environment.

## Use cases
- run a pretrained AI agent that plays Minecraft
- reproduce the VPT paper's behavioral cloning results
- fine-tune video-pretrained models on contractor demonstration data
- train agents to act by watching unlabeled gameplay videos
- participate in the MineRL BASALT competition
- experiment with inverse dynamics model labeling of videos

## When to choose
- you want to run or fine-tune OpenAI's pretrained Minecraft agents
- you're researching learning actions from unlabeled video
- you need a baseline for MineRL or BASALT competition work

## When to avoid
- you need a general-purpose video understanding or video generation tool
- you want a maintained library for domains other than Minecraft/MineRL
- you need Python 3.10+ compatibility with pinned dependencies

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, reinforcement-learning, video-processing, simulation
- domain: machine-learning, deep-learning, reinforcement-learning, gaming-tools, computer-vision
- platform: python, windows
- tags: video-pretraining, behavioral-cloning, minecraft, minerl, imitation-learning, openai, pretrained-models, linux, macos

## Member repositories
- openai/Video-Pre-Training (main) score 41

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:29.870777+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:30:38.722711+00:00, confidence not recorded.
  - readme: https://github.com/openai/Video-Pre-Training (fetched 2026-08-28T04:05:29.870777+00:00, sha 64a157386d6e)
- Data as of 2026-08-30T08:39:29.467469+00:00.