# SandAI-org/MAGI-1

MAGI-1: Autoregressive Video Generation at Scale

Repository: https://github.com/SandAI-org/MAGI-1
Canonical: https://ross.abutalabs.com/products/magi-1
Homepage: https://sand.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: autoregressive, diffusion-models, video-generation
Last push: 2026-06-17T07:06:35+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 88, release rhythm 35, longevity 35
- inputs: {"age_days": 499, "days_push": 77, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3772, forks 240 (observed 2026-08-28T04:08:18.151920+00:00)

## What it is
MAGI-1 is an open-source autoregressive video generation model from Sand.ai, released with Apache-2.0 licensed code and weights. It generates video via chunk-by-chunk autoregressive diffusion, supporting text-to-video, image-to-video, and streaming/continuation generation.

## Use cases
- generate videos from a text prompt
- animate a still image into video
- extend or continue an existing video clip
- run a video generation model on my own GPUs
- research autoregressive video diffusion architectures
- build an AI video creation app

## When to choose
- you need open weights for video generation you can self-host and fine-tune
- you want streaming or autoregressive video continuation rather than one-shot generation
- you have sufficient GPU resources and want full control over the model

## When to avoid
- you only need a hosted API without managing GPUs - use the Sand.ai platform instead
- you lack high-end GPU hardware, as inference is resource-intensive
- you need audio-video joint generation - that is the newer MAGI-2/GAGA line

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, video-processing, llm-inference
- domain: deep-learning, artificial-intelligence, image-processing
- platform: python
- tags: video-generation, autoregressive-model, diffusion-models, text-to-video, moe, open-weights, video, gpu, linux, docker

## Member repositories
- SandAI-org/MAGI-1 (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:18.151920+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 2026-08-29T18:27:52.385582+00:00, confidence not recorded.
  - readme: https://github.com/SandAI-org/MAGI-1 (fetched 2026-08-28T04:08:18.151920+00:00, sha 67d9126f64cc)
  - homepage: https://sand.ai (fetched 2026-08-29T09:22:43.834574+00:00, sha 3b85a062a625)
  - site_page: https://platform.sand.ai/docs (fetched 2026-08-29T09:22:43.843954+00:00, sha c75de6bea7de)
  - site_page: https://sand.ai/about-us (fetched 2026-08-29T09:22:43.845794+00:00, sha da1f17c0f310)
- Data as of 2026-08-30T08:39:29.467469+00:00.