like a black and white storyboard before creating a video
lighting options
year
mood , comedy,drama,action,etc
country
place
planet
This should be done via prompting.
My perspective is that the user works out the requirements for the video step-by-step or creates a separate prompt for a LLM like GPT-4 and runs this in advance before tasking the, likely expensive, video model. This guideline could then be reapplied for additional scenes.
Following this process would end, or start, with a clearly defined guideline for how the video should look like before working on the details.
It would be fantastic to have a cheap and fast video preview of the video before starting the generation process.
1 Like