org card
Browse files
README.md
CHANGED
|
@@ -12,8 +12,7 @@ and scene together in one shared 4D space — the foundation under Meme Maker, P
|
|
| 12 |
and the Viggle API, used by 50M+ people.
|
| 13 |
|
| 14 |
**This organization is where we open-release models.** Weights, inference code and a runnable demo
|
| 15 |
-
Space for each one.
|
| 16 |
-
[MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3) and load no text encoder at inference.
|
| 17 |
|
| 18 |
## Open releases
|
| 19 |
|
|
@@ -63,14 +62,3 @@ generate. Takes of 3–10 s at 24 fps.
|
|
| 63 |
[Discord](https://discord.com/invite/viggle) ·
|
| 64 |
[X](https://x.com/ViggleAI) ·
|
| 65 |
[YouTube](https://www.youtube.com/@ViggleAI)
|
| 66 |
-
|
| 67 |
-
## Before you download
|
| 68 |
-
|
| 69 |
-
The weights are licensed under the **MiniMax H3 Community License**, whose Applicable Territory
|
| 70 |
-
excludes the European Union, the United Kingdom, the Republic of Korea and the United States —
|
| 71 |
-
read it on the model page before you download. The code we ship alongside is Apache-2.0, and
|
| 72 |
-
Meridian additionally needs Meta's VGGT-Omega, obtained separately under a noncommercial licence.
|
| 73 |
-
|
| 74 |
-
Both models run on one high-memory GPU as released. Community work on quantization and offloading
|
| 75 |
-
toward consumer cards is welcome — the architecture is unchanged from MiniMax-H3 and the text
|
| 76 |
-
encoder is never loaded.
|
|
|
|
| 12 |
and the Viggle API, used by 50M+ people.
|
| 13 |
|
| 14 |
**This organization is where we open-release models.** Weights, inference code and a runnable demo
|
| 15 |
+
Space for each one.
|
|
|
|
| 16 |
|
| 17 |
## Open releases
|
| 18 |
|
|
|
|
| 62 |
[Discord](https://discord.com/invite/viggle) ·
|
| 63 |
[X](https://x.com/ViggleAI) ·
|
| 64 |
[YouTube](https://www.youtube.com/@ViggleAI)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|