Home Artists Posts Import Register

Downloads

Content

For those who are generating... there are no available prompts because nothing worked (as expected) so I had no other option than do the TRIPLE-GENERATION method. What it is? Listen carefully:


You have a bad model (or bad LoRA style) = everything is full of artifacts and glitches

1) Generate as normal TEXTtoIMG including all LoRAs + 20-50 (images of the same thing)

2) Pick the top 5 and scale them up (2-2.5x)

3) Import one of them to IMGtoIMG and change 'Checkpoint' to a different one (visually similar one, not completely different). Import also all prompts BUT without all LoRAs only text.

4) Generate between 40-60% influence (don't scale up) [test multiple]

5) Import the best result into 'Inpaint' and select only the face with a mask. And generate the face for the best quality (if necessary).

WHY am I saying it?... well, there are only a few pictures but each of them took me 20+ minutes to make. AI is good but SOMETIMES NOT xD

Files

Comments

No comments found for this post.