Please post the first picture directly here in the Model Gallery. This is the best way to support creators :)
Hello ♥
for whatever reason you want to show me appreciation, you can: ❤️Ko-Fi❤️
_________________
This was my first images back in 2022 with Midjourney with the prompt:
green spongebob squarepants with tons of oreo
Imagine, this was my first ever generated image with AI and this was the first “prompt” I had in my head 🤣
Now, with this Lora and ForRealXL this is the result
This Lora was a test with a new captioning method that delivers much better results than expected.
So you can try it out too.
This Lora was trained with 34 Midjourney V6.1 images. (Version 0.1)
Settings
0.3 - 1.0
Trigger (v0.2+)
aidmaMJ6.1Trigger (v0.1)
midjourneyv6.1Hope you get a lot of beautiful results ♥
Description
v0.1
you will get more details combined with the typical Midjourney Style
FAQ
Comments (27)
if something is early access does it show up in searches when looking to add a lora?
that's a good question, i'm not sure about that
Thank you for making this LoRA 🩷! It makes my images way better 🤩Keep up the amazing work 🩷
Thank you so much ♥
care to go into more detail about the captioning method used?
its a mix between wd14, an own prompt and Joy.
@AIDigitalMediaAgency Joy is the secret
There are a lot of people who want to help
GJ
Great aesthetic! Well done.
♥
Absolutely amazing lora! good job!
♥
Really cool to play with it,congrats!
Thank you ♥
good results, thank you!
Guys, just out of curiosity, I came from SD1.5 and moved straight to FLUX. Obviously mind-blown, however, in my (limited) experience, I have to admit that SDXL is probably more accurate to prompt. On the other hand, FLUX seems to produce way higher quality images. Anybody got some thoughts of dis-/agreement?
Flux is way better than everythin else. The step from SD 1.5 to FLUX is HUGE!
@AIDigitalMediaAgency Sure, generally speaking. However diving into aspects... Wouldn't you agree prompt accuracy is better on SDXL? I hosted a bounty for a 4 armed person. It's not impossible with Flux, but you have to be a whole lot luckier than somebody on SDXL. Most if not all of the entrees leveraged SDXL.
https://civitai.com/bounties/3904/search-for-goros-girlfriend
FLUX is the highest quality - bar none. However... Flux is absolutely NEW and there are a number of things it falls short on or simply can't produce natively (artists/art styles). That's to be expected. I have SDXL output images Flux can't come close to reproducing (Even with Forge's, Flux enabled img2img).
Patience - Fine Tuned Flux models will start coming out and bridge those gaps. 2 or 3 months out, things will be unreal.
@dociler There are several aspects to "prompt following". The "4 arms" example is a "concept/idea", and a model will not be able to render it unless the concept/idea was trained into it. Most of the SDXL examples you see are probably generated using some sort of PonyV6 model, which probably has that "4 arms" concept trained into it. Base Flux probably does not have such an unusual concept/idea baked into it.
But "concept/idea" prompt following is a solve problem: you just train a LoRA for it. What Flux provides is a more sophisticated, difficult type of prompt following, where you describe an image in great detail, and have the model render it properly.
Here is an example: https://civitai.com/posts/5965274
It is basically impossible to do it with any SDXL, with or without LoRA. The only way to it is some complex use of ControlNet, combined with manual inpainting, and img2img.
It loves putting white blotches over everything at 1.0 setting on my setup for some reason.
Do you have an example?
Try the SGM Uniform or Beta scheduler
Feelsl ike Im getting vertical lines, any idea why?
No idea what you mean, do you use Flux dev.?
Do you have an example?
@AIDigitalMediaAgency same im using this lora with flux get vertical line on edge image
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.