Pinned post
My new short “test” film for my cinematic AI workflow.
——
The days of “what model did you use?” is mostly gone, i.e. just sharing the prompt and model is no longer enough to explain how you made an AI video. The prompt and the model is just the last step of a very long workflow.
At least that’s what I’m finding as I try to figure out how to make AI video models do what I want them to do.
It is a sign that AI video models are becoming less random and more directable, which was not the case a year ago. With that ability, you need to learn and develop the right tools to direct what you want in more and more sophisticated ways. You can do this by building a human workforce, akin to the classic production teams in Hollywood which many are doing, or you can use LLM models that are becoming increasingly more capable not only in generation but in being able to see and analyze images, video, and spatial schematics like floor plans or 3D models.
I am in the LLM category and I’m pushing it until I find a hard boundary of what the limits are. I have not reached that boundary yet and that boundary is continually being pushed out as new models are released.
It is quite an extraordinary time we are living in and I hope you enjoy my latest “test” film as I continue to improve my cinematic AI workflow to hopefully make it good enough to be able to do a feature length film one day.
——
The days of “what model did you use?” is mostly gone, i.e. just sharing the prompt and model is no longer enough to explain how you made an AI video. The prompt and the model is just the last step of a very long workflow.
At least that’s what I’m finding as I try to figure out how to make AI video models do what I want them to do.
It is a sign that AI video models are becoming less random and more directable, which was not the case a year ago. With that ability, you need to learn and develop the right tools to direct what you want in more and more sophisticated ways. You can do this by building a human workforce, akin to the classic production teams in Hollywood which many are doing, or you can use LLM models that are becoming increasingly more capable not only in generation but in being able to see and analyze images, video, and spatial schematics like floor plans or 3D models.
I am in the LLM category and I’m pushing it until I find a hard boundary of what the limits are. I have not reached that boundary yet and that boundary is continually being pushed out as new models are released.
It is quite an extraordinary time we are living in and I hope you enjoy my latest “test” film as I continue to improve my cinematic AI workflow to hopefully make it good enough to be able to do a feature length film one day.
2:38
