Generative AI - Flux (text to image)

[freezefantasy]

The bar has been raised for consumer-grade image generating AI.

Here's a couple URLs to images made out of the box with very simple prompts (no customisation/fine tuning or loras). Seems like we can't post media on here anymore?

I've posted more examples and details on the discord channel for anyone interested.

https://fal.media/files/koala/Gs3-rFws5PtU3xy7VeiFq.png
https://fal.media/files/monkey/QtYcSUhX ... cDKDLW.png
https://fal.media/files/koala/Mch8O0qhRXU_MynQCsh5F.png
https://fal.media/files/lion/Y1Fb7oKnBwQCJLhUqwRtp.png
https://fal.media/files/zebra/DuwCCc5zSP2WmYXLM0XmO.png
https://fal.media/files/panda/p2awTmdaSvnmv_8z4pSln.png
https://fal.media/files/elephant/R4-mWy ... k71-Qs.png
https://fal.media/files/koala/Gs3-rFws5PtU3xy7VeiFq.png

The model is called Flux, and you can try it on this playground: https://fal.ai/models/fal-ai/flux/dev

For a more premium and NSWF capable version, use Flux Pro: https://fal.ai/models/fal-ai/flux/pro

[Freeze_]

Flux is great. A big improvement. I've been having fun with it this past week

[neisy03]

It cant seem to produce this kind of image with the same prompt though.

[HypnoGold]

It cant seem to produce this kind of image with the same prompt though.

Can you share your prompt for this? I need help fine tuning mine

[freezefantasy]

[Xban]

Awesome find, one of the better ones I've seen

[freezefantasy]

Black Forest Labs has released a super easy way to train your own adapter models (LoRas) to feed into a Flux image generation. You just need to provide reference pictures, a 'trigger word' and all subsequent images that include this word will utilise the styling or face / body of your reference images. Here's an example of a reference girl I wanted to generate more photos from.

And here is the example of what type of outputs can be achieved by adding this adapter + including the trigger word 'ywx' into a simple prompt ("girl poses naked as mannequin in a store display ywx")

Adapter Training here ($2 / adapter approximately): https://fal.ai/models/fal-ai/flux-lora-fast-training
Flux AI Lora-adapted Generation ($1 for 20-30 HD images): https://fal.ai/models/fal-ai/flux-lora

No output is perfect, so you do need to generate like 5-10 images to get one that might be passable (fingers, toes and limbs return as challenges when using strong adapters). But it's crazy how easy it was and relatively affordable. As you can see nudity is allowed but you have to do it via the API instead of on the website.

[Samsung]

I've made over 300 images on my own private computer using celebrity LoRA's from Civitai.com and experimenting with prompts. I used Stable Diffusion Forge webUI on a Macbook Pro M1. It's free and stays on your machine.

[Samsung]

[neisy03]

00007-997318268.png
00009-2860901847.png
00000-2648280953.png

may I suggest to generate images like these?

these are easy to generate in bing, I tried generating them with same prompts in other free AI sites, but they don't generate same qualities of images.
im just curious if your AI can do the same quality.

[Samsung]

I'm using the Flux model with SD Forge so yes it can do the same quality or better. Flux is one of the best models out there. Just YouTube how to install Stable Diffusion Forge and you can create images for free on your computer.

[ensignronald]

I started posting my AI stuff on DeviantArt. It's mostly chubby women turned to stone/cement encased/turned to metal/frozen solid. Most were made with CivitAI (the nude ones) and Google Labs AI (which recently relaxed a lot of their intense restrictions).
https://www.deviantart.com/haaardtimes

[Leem]

Recently I've been using Perchance, which requires no login, and allows unlimited generations, no censorship and numerous models. It admittedly retains many drawbacks from earlier programs, but the results can be good with a little patience. I haven't posted many of these on line yet, but sooner or later I'll probably put a page or two (dozen?!) of the best ones on my site.

[mannequinlover]

Speaking of generative photo AI. Give this site a go…. No login required. I’d just make sure you’re connected to a VPN when using it, but OMG!!!!! It’s absolutely fu*king amazing!!! I don’t think they have a lot of the guard rails like some of the other sites. Man you can spend hours upon hours generating some of the best photos you can imagine!!!

The site is: https://flatai.org/ai-image-generator-free-no-signup/

[mannequinlover]

[mannequinlover]

[suitedmannequin]

What prompt exactly did you write to have these photos ?

[mannequinlover]

If you’re referring to the photos I just posted, basic idea is:

“Woman mannequin with shiny flesh toned skin, brown hair, and wearing jean shorts and a swimsuit top”

That’s the basic idea. You can toss sentences after that like “she is sitting on a table” etc to fill in the rest. Try “she’s doing yoga” as an action … wow the images are absolutely eye popping!!! . I was completely floored when I first saw how flat ai produced images from that prompt!!! …. Be careful though, it’s highly addicting, as I’ve spent an obscene amount of hours on this when I first discovered it not too long ago lol! Like anything, I’m sure the boredom will set in, but for now it’s all so mind boggling! I’m like a kid on Christmas every time I hit the generate button!

Cheers!

[yendis]

my first attempt with FLux
cheers
yendis