The generation process itself doesn't follow the prompts I imagine.
The generation process itself doesn't always follow my expected prompts, but I use it for upscaling images output in GPT format. It's the best for upscaling. For creating videos from images, the prompts must be set up in a way that doesn't damage the original image; simply generating images or specifying movement will result in a loss of quality. Therefore, image generation from prompts isn't ideal. I feel it still has a long way to go to justify paying for its generation. Furthermore, for those using GPT on a subscription basis, there's no point in heavily paying for this AI. The price of video creation and the output don't justify the cost. I hope for future improvements.