Description
Trained on 39 images.
FAQ
Comments (5)
hi, what are your training settings? your loras look excellent!
I use Kohya Dreambooth with pretty much all the default settings: https://github.com/Linaqruf/kohya-trainer
And try to find really high def images to train with.
The LOFI model I use to render is one of the biggest reason they look good: https://civitai.com/models/9052/lofi
@chairfull Thank you for your great LORA. Have you ever tried making them using the Checkpoint Realistic-Vision? It also creates a "RAW, Analog Style" image effect ("Lo-Fi effect"), but the portrait photos created look so much more Realistic. I have been using your LORA using "Realistic Vision".
@maxhitman I've been trying it for some models, but can't get the same level photorealism:
https://civitai.com/models/28421/charlie-hunnam
https://civitai.com/models/30865/jeffry-donovan
Maybe I need to peek at other peoples prompts and find better tokens.
@chairfull Since you create such nice Lora for us, I will share you a secret "Prompt"
for making better Photo-realistic images using Realistic-Vision, and also many other photo-style Checkpoints.
- The positive "Prompt" has to be separated by commas and each Section of the prompt describes each area of the image which should be worked on by the A.I. - First is to describe the type of image you want to see, then the person, then the body type, then the photo style and effects , then the background or scene the person is in, then you describe other camera type, lens type, other effects, making sure the A.I. knows that you want a high-resolution photo image and to finalize, you use the words, "award winning" and/or "masterpiece". This is the secret to qa better A.I. written prompt
...this is for a photo image of a man... (Change the words according to your needs or what you NEED to see in the photo and also keep the CAPITAL LETTERS as seen in my written Prompt below , to make that
word more important.)
.................................................................................................................
...this is for a photo image of a man... (Safe -For-Work)...
Full Body Portrait of 30 year old Film Actor CHARLIE HUNNAM man with Long Blond hair style, Blond short beard style, Detailed face, (Smile), (perfect eyes), (highly detailed skin:1.1), Perfect Athletic Body, wearing a shirt and Jeans pants , Modelshoot style, Professional Photography, Soft Lighting, PHOTOREALISTIC, Realistic, standing in a dark studio background, blurred background, RAW, analog, sharp focus, 8k, high resolution, DSLR, high quality, Fujifilm XT3, film grain, award winning, masterpiece
.................................................................................................................
...this is for a photo image of a woman... (Safe -For-Work)....
Full Body Portrait of 25 year old Singer TAYLOR SWIFT with blue eyes and Long Blonde Hair Style, (Perfect eyes), (Eye Makeup:1.1), Detailed Face, (Smiling face expression showing teeth), (highly detailed skin:1.1), (perfect breasts:1.1), Perfect Slim Body, Modelshoot style, wearing an elegant short dress designed by Versace, Professional Photography, Soft Lighting, PHOTOREALISTIC, Realistic, Sunset, foggy, standing in a forest field with wild flowers background, blurred background, volumetric fog, RAW, analog style, sharp focus, 8k, high res, DSLR, high quality, Fujifilm XT3, film grain, award winning, masterpiece
.................................................................................................................
If you check out my LORA of Sigourney Weaver-Ellen Ripley, I have noted some other Tutorial Tips I wrote and some secret "trigger words" to use in your written Prompts (which can be used with many Checkpoints) for a more realistic style A.I. create-photograph. Also works well when using Anime-style Checkpoints with a
3-D (or 2-D) style to the images.
Cheers
MAX