how to buy access?
how to get access?
Don't pay for this poor implementation...not worth it.
Honestly at this point would rather spend money to buy credits for seedance 2.5 than pay access for this. It's weird that a basic 0.1 alpha version wasn't even released. Nobody is posting videos anywhere to even encourage other's to donate. Hopefully lightricks is working a new next generation video model that is more convenient to use than h3.
Honestly at this point would rather spend money to buy credits for seedance 2.5 than pay access for this. It's weird that a basic 0.1 alpha version wasn't even released. Nobody is posting videos anywhere to even encourage other's to donate. Hopefully lightricks is working a new next generation video model that is more convenient to use than h3.
Its out now. And no further donations were closed long ago and none have been accepted since. Minimax H3 is terrible to train, but some say its gotten better. But when this was started, things hadnt matured properly. Whoever releases a new video model with real documentation and a base to train on, is going to win this race for sure. Minimax has largely ignored its community after H3 was released. We didnt get the 2k refiner, training documentation nor a fixed ref model and as far as I know there's been no mention of H3.1 or anything of the sort. While Flux3 Video appears to be MIA.
Honestly at this point would rather spend money to buy credits for seedance 2.5 than pay access for this. It's weird that a basic 0.1 alpha version wasn't even released. Nobody is posting videos anywhere to even encourage other's to donate. Hopefully lightricks is working a new next generation video model that is more convenient to use than h3.
Its out now. And no further donations were closed long ago and none have been accepted since. Minimax H3 is terrible to train, but some say its gotten better. But when this was started, things hadnt matured properly. Whoever releases a new video model with real documentation and a base to train on, is going to win this race for sure. Minimax has largely ignored its community after H3 was released. We didnt get the 2k refiner, training documentation nor a fixed ref model and as far as I know there's been no mention of H3.1 or anything of the sort. While Flux3 Video appears to be MIA.
after my terrible experience two attempts at lora training for minimax h3 and wasting nearly $400 in runpod credits. I'm not surprised by the sentiment that this finetune was terrible to train. As body horror and limited as ltx 2.3 was, at least you got a sense that it learned something from lora training and there visual changes occurring when the lora weight is applied. with Minimax h3, it never feels like it's learning anything at all no mater the step count, learning rate or rank settings.
I had a blast creating LoRAs for MMH3. No idea what y'all talking about.