-no captions needed during training (experimental)
-trigger words are to help guide your images but do not correspond to the weights of the LoRa


Did some test with/without the trigger words (with the LORA at strength 1):
1- Just the LORA seem to stay close to the "original" image appearance with a little grainy effect.

2- Using the trigger words after a few steps keep the same appearance as "just the LORA", but with a better effect.

3- Using the trigger words for a few steps give a result similar as using the trigger words all the way, but with less grainy effect.

4- Using trigger words for all the steps, gives the expected grainy effect but can heavily affect the original image.
5- No LORA, for comparison.

2 and 4 seem to give the best results, but the other can still be useful.