Anonymous
8/21/2025, 5:11:09 AM
No.106331383
>>106331275
Basic masking with regional prompt is much better than polluting everything in a single text input box with a custom, naturally unintuitive prompting format of "character:a - short hair, smile" or anything like that which also requires retraining of all image gen models, captioning models, and prompting guides
You will always need to use regional prompting if you want specific things in a very specific place, and the rest can be taken cared for with basic language with you speaking to an LLM that can help prompt the model for now until we get natively multimodal image gen.
Basic masking with regional prompt is much better than polluting everything in a single text input box with a custom, naturally unintuitive prompting format of "character:a - short hair, smile" or anything like that which also requires retraining of all image gen models, captioning models, and prompting guides
You will always need to use regional prompting if you want specific things in a very specific place, and the rest can be taken cared for with basic language with you speaking to an LLM that can help prompt the model for now until we get natively multimodal image gen.