cinemasetfreeModels, settings and LoRAs
Each design picks an image model and a text model, exposes that model's advanced parameters, and can attach LoRAs for fine-tuned control of a character, style or concept.
Two models, two jobs
The text model is what Describe uses to write a visual description from what the project knows. The image model is what Generate uses to render it. They are chosen independently because they are good at different things and priced differently; the model catalog has the numbers.
Advanced settings are per model, because a parameter that exists on one model doesn’t exist on another. What each one actually does is covered in Model Settings.
LoRAs
A LoRA is a small trained add-on that teaches a base model one specific thing, a face or a style or a concept, without retraining the model. Attach one to a design and it applies to that design’s generations, with a weight you control.
Sources, weights and combining several at once are covered in the LoRA guides.
Generation spends credits. The estimate shown before a bulk generate is worth reading, and a design change that invalidates a lot of existing images means regenerating them at full cost, so settle the look before you generate the whole cast.