“the easier or the faster part in generating this is actually in creating the voice and the hard part is it takes many minutes to actually process and generate the video with the right lip movement to match the text that you've entered and so rather than showing you something you know lips moving that is off from what you what you've uh put in they first show you just kind of a blurry version with the audio so you can get a sense of like what it's going to be like then you click generate here and then that you know it says 12 minutes right here is how long it's GNA take” — Host

“trading off um basically uh Fidelity for immediacy and basically putting the human kind back in the loop because if it was just a generate button right we would wait for 12 minutes uh figure out that something is not quite right and then kind of like you know give the machine a new prompt and wait until it comes back so this is a really uh clever trick uh to really kind of create this iterative human machine collaboration interface um” — Rafael Shad