Midjourney has been working on a video model for a while now. While we have known that, we didn’t get a chance to test it till recently. The Midjourney V1 video model is an image to video model that generates 4 x 5s clips that can be extended up to 20 seconds. It has low and high motion options for auto and manual video modes. In manual video mode, you can guide what you want the model to do.
As MJ explains, video cost 8x GPU time compared to regular images. Pro and Mega plans can generate videos using Relax Mode. You don’t necessarily need to use parameters through the web interface but -motion low, –motion high, and –raw models are supported. You can generate videos using Midjourney or external images. Midjourney videos are generated in 480p quality.
[HT]
