Minimax H3 doesn't need long. It depends on the duration and the resolution of course. I also have a RTX 5080 (64GB RAM) and 40-60 mins still sounds pretty long for me.
The NVFP4 text encoder seems to be the best one when it's about speed and memory consumption.
I think Maestro only displays the pure denoising time. On top of that time goes text encoding and model loading. But this is what Maestro displayed on my latest run for example. I used the 4 Step model which has an amazing quality.
Minimax H3 doesn't need long. It depends on the duration and the resolution of course. I also have a RTX 5080 (64GB RAM) and 40-60 mins still sounds pretty long for me.
The NVFP4 text encoder seems to be the best one when it's about speed and memory consumption.
I think Maestro only displays the pure denoising time. On top of that time goes text encoding and model loading. But this is what Maestro displayed on my latest run for example. I used the 4 Step model which has an amazing quality.