Project by AiArtLab
Thank you so much for the feedback and examples of your work! It's very motivating.
@_Wizz_ Provided free discord server for us with generations ability and you may test some new versions here: https://discord.gg/HpeGYm9Xae
Pls support my trains at Boosty or Patreon:
https://www.patreon.com/recoilme
Tips and Tricks
Use low CFG (guidance scale)
Euler are best for this model
License CC-BY-NC-4.0
Please ask me about commercial use!
Contacts
My 1.5 models
Description
finetuned on posters, cinematics, art, photos (including nsfw) and image with texts (a little)
FAQ
Comments (5)
I was looking at your Aesthetic dataset on Huggingface, and just so you know, ALL the descriptive captions for porn images are completely wrong. I think the Vision model you used is not capable of recognizing porn at all.
You right, thats old dataset, i use another captions with moondream2 + wdtagger, example
A woman in a black lace dress is kneeling on a patio, her body partially hidden by a white cushion. The patio is furnished with a wicker chair and a wooden bench, and a potted plant stands nearby. The woman's long dark hair cascades down her back, and her eyes are focused directly on the camera. 1girl, solo, breasts, high heels, cleavage, leotard, all fours, long hair, couch, looking at viewer, black footwear, hair over one eye, brown hair, brown eyes, plant, photorealistic, ass, large breasts, black leotard, lips, indoors, potted plant, black hair, full body, long sleeves, medium breasts, curly hair, carpet
but i dont train on porn
@recoilme Hello, could you share with me the image number in your train dataset? all the images are photos or synthetic or hybrid? I think you have much more images than 2k. thx
@GuardSkill its 3k images, photos, art, cinematic, texts and so on. I dont need many for finetune, but i select every image manually
thanks for your friendly reply