Quote:What i am tring to do is simply let the model handle the resizing, i dont want hybrid to touch it. So if i add two x2 models for example a 852x480 video, i want the first model to make it 1704x960 then the second model to make it 3408x1920 and thats it thats the final video nothing else. There is no target resoultion, the target is what the last model gets the video to.The target is always what you configured under "Crop/Resize->Base->Resize->Picture Resize" if that corresponds with what your model will output, Hybrid will not do the last step with the (adjustment) resizer.
Quote:Sorry i should have just mentioned that i am using VSMLRT, since VSGAN has no tensorrt support in Hybrid.Yes, you should state details. The reason for not offering TensorRT with VSGAN (https://github.com/rlaphoenix/VSGAN) is that VSGAN has no option for it.
=> I will look into adding a sub option to 'Multi' to not adjust the resizing between the models after work.
There also seems a bug in VSMLRT + Multi. (will fix that too)
Cu Selur