headchem
commented
2 years ago Defer this to a later phase. Ideally, increased training data will result in high quality stories generated from scratch with a single button press. The phased approach is designed around assumed failure to give user something usable.
To further simplify interface, maybe after logline slides/pages, an entire story is generated? Or at least an option to generate the entire thing? Instead of requiring user to go page by page through each Sequence, which feels tedious. Could group Sequences into 3 Acts, even though they still generate one-at-a-time behind the scenes, so user only has to click 3 times to generate entire story.
For each generated Act, user can select a sequence to lock, which locks all previous sequences, and regenerate remaining unlocked sequences.