OK to use in the USA?
The license doesn't grant use in the USA and I am not sure if it's an oversight or not. I am only wanting to use it locally but want to respect the license.
minimax famously stole their training data from all of these places and others (google it) . For personal use i would not feel too bad about it. Nobody else does.
That may also be part of "why", to avoid legal trouble in those areas with said original data creators (think Hollywood...). Minimax would likely look bad in a legal copyright attack.
They basically clear themselves of all accountability in these places, by saying "It can't be used there".
But that also makes it VERY unlikely that they sue you there.
minimax famously stole their training data from all of these places and others (google it) . For personal use i would not feel too bad about it. Nobody else does.
That may also be part of "why", to avoid legal trouble in those areas with said original data creators (think Hollywood...). Minimax would likely look bad in a legal copyright attack.
They basically clear themselves of all accountability in these places, by saying "It can't be used there".
But that also makes it VERY unlikely that they sue you there.
You've got to be kidding me. First of all, where's your evidence? Second, if that's the case, why are American AI video generation models so damn terrible? The best video generation models in the world right now are all Chinese. If Hollywood has such an incredible wealth of data, why isn't the US training its own models?
But by all means, I encourage you to keep speaking up and make sure MiniMax sees your comments. That way, when they release the much larger H3.1 model, they might decide not to make it freely available to you anymore. You'll be stuck using that outdated version you claim was trained on stolen data for the rest of your life. Haha!
minimax famously stole their training data from all of these places and others (google it) . For personal use i would not feel too bad about it. Nobody else does.
That may also be part of "why", to avoid legal trouble in those areas with said original data creators (think Hollywood...). Minimax would likely look bad in a legal copyright attack.
They basically clear themselves of all accountability in these places, by saying "It can't be used there".
But that also makes it VERY unlikely that they sue you there.
I took a quick look at the image-to-video arena, and it was dominated by Chinese models. Several of the competing models were actually derivatives fine-tuned on top of MiniMax.Putting aside whether your arguments about the training data hold up, I simply hope some American companies will open-source better video models instead of just focusing on making money.
You've got to be kidding me. First of all, where's your evidence? Second, if that's the case, why are American AI video generation models so damn terrible? The best video generation models in the world right now are all Chinese. If Hollywood has such an incredible wealth of data, why isn't the US training its own models?
But by all means, I encourage you to keep speaking up and make sure MiniMax sees your comments. That way, when they release the much larger H3.1 model, they might decide not to make it freely available to you anymore. You'll be stuck using that outdated version you claim was trained on stolen data for the rest of your life. Haha!
why don't you ask Disney and so on? Haha.
Funny how these regions are those with quite firm and enforced copyright laws, isn't it?
I'd love to see what would happen if someone there used it commercially and they try to sue.
Also has it occured to you that that is a big reason why there are so few non chinese image/video models?
Btw look at Fibo2, it was trained "only on licensed material". So as a result nobody cares about it. (It's ok at 1MPx but deteriorates on higher resolutions.