I am a film maker/ storyteller. I have worked in Video and CGA. I excel at writing, set design and editing. I am just getting starting (this week, LOL) learning AI. My goal is to make narrative drama films in AI. I want to be able to control camera placement and movement to create individual shots. I need to be able to “direct” actor’s movements and expressions and hopefully have them deliver the lines I have written. Then I want to be able to take multiple shots into Premiere Pro, edit shot lengths and transitions. Then color correct and color grade. I create music in ENVATO and voice over dialog in REVOICER.
QUESTION: Is this even possible? What software should I start with?
Camera placement, individual shots etc are very doable with Seendace 2.0, Minimax H3 models for example. You describe the shots, including with camera framing and movement in the prompt. Same thing with the lines and expressions. However this almost never works out on first try, think of it as multiple takes when making a movie, and then stitching them together.
I have finished the script. I am just about to start next week on generation. I am looking at Seedance 2.5. Is that a big difference from 2.0?.. BTW I did a course on Udemy. They recommended using ChatGPT to write the script and get prompts. That turned into a disaster. I have always developed my own stories and written the spec script… That works better for my creation process.
My tool allows for script ingest and preparation of project. The script, in standard script format, can be loaded, and then the tool creates scenes, characters and environments listed in script
Extracts dialog, etc.. does a lot to advance the work
Yes I haven’t seen AI yet make good scripts from scratch, I find it good if you have an idea and you iterate with ChatGPT for example. When it comes to Seedance 2.5, to be honest I have been very disappointed personally, but maybe that is because I work on vertical and animated, I do not see any great improvements to Seedance 2.0. Recently I have been having best results with Minimax H3, it continues to amaze me, and I think as it is open source and cheaper, that might be where the future is headed. I mean imagine, you get your results back in less than a minute, sometimes seconds with Minimax H3!