Been following the ai wam ideas for a while, yesterday I decided to have a go myself!
Using https://www.seaart.ai/ which uses stable diffusion for generation. Completely free and user friendly! Was daunting at first, but building from simple prompts, adding prompts from art I liked and suggestions I found, quickly I was able to generate usable images.
I recommend anybody to have a go, and see what they can create.
Model used : Level4, LoRa used : Zovya's Wet Hair, at 1.5 weight
I've added some iffy pics as the wet clothes effect worked well on them.
renfield9 said: Been following the ai wam ideas for a while, yesterday I decided to have a go myself!
Using https://www.seaart.ai/ which uses stable diffusion for generation. Completely free and user friendly! Was daunting at first, but building from simple prompts, adding prompts from art I liked and suggestions I found, quickly I was able to generate usable images.
I recommend anybody to have a go, and see what they can create.
Model used : Level4, LoRa used : Zovya's Wet Hair, at 1.5 weight
I've added some iffy pics as the wet clothes effect worked well on them.
Certainly very impressive results.
For those of us starting out, it would be useful to know the prompts you used for maybe just one of these stunning images. Also, did you discover any banned prompts with seaart, as is certainly the case with Bing ?
Thanks for all the feedback guys. These were all basically the same prompt, apart from adding standing in water to some, and tweaking dress style
30 year old cute flirty brunette woman in ((wet midi dress)), wet hair, standing in water, soaked, braless, pantyhose
Negative prompts are super important to avoid cthululu monsters. My default negative prompt dump is below.
cut off, bad, boring background, simple background, More_than_two_legs, more_than_two_arms, (skinny), (ribs), (bony), (blender model), (fat), ((((ugly)))), (((duplicate))), ((morbid)), ((mutilated)), [out of frame], extra fingers, mutated hands, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), ((ugly)), blurry, ((bad anatomy)), (((bad proportions))), ((extra limbs)), cloned face, (((disfigured))), out of frame, ugly, extra limbs, (bad anatomy), large breasts, gross proportions, (malformed limbs), ((missing arms)), ((missing legs)), ((extra arms)), ((extra legs)), mutated hands, (fused fingers), (too many fingers), ((long neck)), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist's name, asian, unnatural pose, render, anime, cartoon, painting, thong, airbrush, (3d, render, cgi, doll, painting, fake, cartoon, 3d modeling:1.4), (worst quality, low quality:1.4), monochrome, child, deformed, malformed, deformed face, bad teeth, bad hands, bad fingers, bad eyes, long body, blurry, duplicate, cloned, duplicate body parts, disfigured, extra limbs, fused fingers, tattoo, extra fingers, twisted, distorted, malformed hands, mutated hands and fingers, conjoined, missing limbs, gross
I discovered last night that using the model - am i real, means no need for wet hair LoRa's, I've attached some attempts from last night. You may recognise the ladies....
I have discovered it is easier to work with the AI rather than against. Start with something easy to render wet, like satin or chiffon, and build it up from that.
Good work! 4th one on the yellow is the best one. I think to make some of them better you need to give the girls a few more imperfections like freckles or moles.
I really like them on one hand but feel slightly disappointed its not a real person or a real image on the other hand. The fact they are almost 'too perfect' reminds me of Sirens (Especially with the wetlook) makes me expect the image to change into a monster any moment!
Andytac said: Following renfield9's hints, some of my first attempts with Seaart...
i try it also , en came across exact same pictures as posted before . i have a feeling this site is not using AI but pre programmed pictures you can choose from by using a promt . for an exaple , if i have a foto of a woman in een blue dress , and i want to change the color of the dress , results are a compleet diverend woman , or the foto does not change at al . it is not posible that we generate pictures with diverend promps en the pictures turnout to be exact the same maybe the pictures are origin made by ai but not on this site .
Hi SatinWAMBossChris. Renfield9 's original list of negative prompts is worth copying & pasting to prevent - at least to an extent - deformities in the output.
SatinWAMBoss Chris said: None of mine looked as realistic. Also, many of them had extra fingers or oddly placed body parts.
It's all down to which model you use. The default ones are aimed more at anime and cgi styles. Level4 and Am I Real are the two I use most as they have realistic lighting, skin, hair etc.
@gebakje I assure you these were all created by ai. What you aren't seeing is the literal hours and dozens of batches of images that didn't work out how I wanted. It isn't going to make the same image each time as it starts from scratch each time. I have noticed like you said styles and faces can often stay similar, but I put it down to that being what matches the prompt most. SD can accept really complex prompts if desired, but I like to stick with "summer dress" and see what happens. If you were to try "long off shoulder linen summer dress with floral detail" you would get a very different image compared to "cute satin summer dress with lace edging". Tweaking the wording each time will get different results. Sometimes wildly different.
SatinWAMBoss Chris said: None of mine looked as realistic. Also, many of them had extra fingers or oddly placed body parts.
It's all down to which model you use. The default ones are aimed more at anime and cgi styles. Level4 and Am I Real are the two I use most as they have realistic lighting, skin, hair etc.
@gebakje I assure you these were all created by ai. What you aren't seeing is the literal hours and dozens of batches of images that didn't work out how I wanted. It isn't going to make the same image each time as it starts from scratch each time. I have noticed like you said styles and faces can often stay similar, but I put it down to that being what matches the prompt most. SD can accept really complex prompts if desired, but I like to stick with "summer dress" and see what happens. If you were to try "long off shoulder linen summer dress with floral detail" you would get a very different image compared to "cute satin summer dress with lace edging". Tweaking the wording each time will get different results. Sometimes wildly different.
that does not explane why i create and find exact the same images as in the post before . no change that you and me use exact the same promts .
Andytac said: Hi SatinWAMBossChris. Renfield9 's original list of negative prompts is worth copying & pasting to prevent - at least to an extent - deformities in the output.
renfield9 said:It's all down to which model you use. The default ones are aimed more at anime and cgi styles. Level4 and Am I Real are the two I use most as they have realistic lighting, skin, hair etc.
my first attempt, using all the info you have been so kind as to provide me with... i need to soak the dress more... but as of now, if you'll excuse me, I think I need to take a cold shower...
SatinWAMBoss Chris said: my first attempt, using all the info you have been so kind as to provide me with... i need to soak the dress more... but as of now, if you'll excuse me, I think I need to take a cold shower...
Stunning work! Try adding tags such as (((drenched))) (((soaked))) (((dishevelled))) (((clinging))) to increase amounts of wetness on clothes.
Andytac said: Stunning work! Try adding tags such as (((drenched))) (((soaked))) (((disheveled))) (((clinging))) to increase amounts of wetness on clothes.
Thank you! Great idea! I will try a few extra prompts and see what else I can do... here are a few others...
Andytac said: Stunning work! Try adding tags such as (((drenched))) (((soaked))) (((disheveled))) (((clinging))) to increase amounts of wetness on clothes.
Thank you! Great idea! I will try a few extra prompts and see what else I can do... here are a few others...
I very much like those pics - what wording are you using in your prompts for the dresses please and have you been able to specify anything which reliably controls the length of them? Have you tried out images with skirts yet?