MAGI-1
Sand AI's Apache-2.0 autoregressive video model that generates video chunk by chunk (24B and 4.5B weights), strong at image-to-video and long, streamed generation.
Visit MAGI-1Free (open source) · Free plan · checked Oct 26, 2026
MAGI-1 at a glance
| Type | Open-source model |
|---|---|
| Developer | Sand AI |
| Starts at | Free (open source) |
| Billing | Open source (free) |
| Free plan | Free plan: Code and weights under Apache-2.0. |
| Watermark | Not stated |
| Max resolution | Not stated |
| Max length | Not stated |
| Commercial use | Not stated. Apache-2.0 licence. |
| API | Not stated |
What MAGI-1 says it can do
- Text to video:
t2v: Text to Video
source - Image to video:
It achieves strong performance on image-to-video (I2V) tasks conditioned on text instructions
source - Extend video:
MAGI-1 further supports controllable generation via chunk-wise prompting, enabling smooth scene transitions, long-horizon synthesis
source - Open weights:
This repository contains the code for the MAGI-1 model, pre-trained weights and inference code.
source
MAGI-1 alternatives
12 tools with the most overlap: CogVideoX, HunyuanVideo, LTX-2 / LTX-Video, Open-Sora, SkyReels, Wan, Genmo, Luma Dream Machine and more. Compare all 12 MAGI-1 alternatives →
Where MAGI-1 ranks
Positions in our lists, from public data only (same grey chip for every position).
- #5 in open-source AI video generators · ranked by public data
- #16 in free image-to-video AI generators · ranked by public data
- #23 in free text-to-video AI generators · ranked by public data
- #25 in image-to-video AI generators · ranked by public data
- #33 in text-to-video AI generators · ranked by public data
- #37 in free AI video generators · ranked by public data
Sources
Checked Oct 26, 2026 on MAGI-1's own site. Prices change; confirm on the vendor's site before you buy. Spotted an error? Send a correction.