HYPERREALISTIC
Hi everyone 👋
Lora features :
High quality images used
Style used (hyperrealistic, hyperrealism, realistic, semi-realistic)
Lora focus on (close-up, portrait, upper body)
Trigger words maybe (hyperrealistic, closeup shot)
Note ⚠️ : You should be aware that positive and negative Prompts completely affect the results, so be careful when writing them.
don't forget LIKE and COMMENT and REVIEW if u love it
Description
FAQ
Comments (21)
@Zoropaton add models in https://www.seaart.ai/ pls
Maybe someone can help me out with this one - because I can't get a character to smile without it being kinda 'off'.
I understand there is a problem with the way characters smile, but you can fix it by mixing it with another lora or decreasing the weight , like this post
Zoropaton Actually I'd say that all of those are good examples of the issue I'm having — they're very aesthetic but in the expression there is something off about all of them. The first one isn't so bad, it could be interpreted as 'serene'; but the others have that 'off' quality.
The eyes are wide, a vacant stare that lacks the subtle movement or emotion you'd expect; especially given the mismatch between the smile and the eyes — the eyes don’t appear to be “smiling” with the mouth. It's vaguely... Psychotic? I guess? A soft smile with a hollow gaze is code for 'you're about to be murdered'.
I've given it a good effort but I can't get it to 'not' do that; with that said outside of this one area it has produced some really nice images.
I see semi realistic
不知道有没有人也有遇到，就是使用stand相关的提示词，就一定会出现一位莫名其妙的机器人🤖，我是了好多次都是这样的结果，后来我才发现，这个机器人的名字应该就是叫做stand🤣，但问题是我该如何避开这个问题使用站立相关的提示词呢...
I didn't understand ur situation, but I think u have a problem with the word " stand "
Try ,standing or stand up, just write it in another way
@Zoropaton yap~, I change stand by other word, everythings is good! I just find it very interesting🤣
@madnessdevilpurg8570 lol, thanks for sharing
二次元里stand有个梗，就是后面有个机器人。要用standing来写动作
单独stand会被AI认为是JOJO里替身的意思，会在身后画一个替身，要用“standing”。
I'm having massive trouble with this lora I love what it can do but I keep getting unasked for blackmasks, googles, gasmask, shirts , uniforms, etc. If the character has an accessories of any kind not added to the prompt it gets one of the above male or female
any clue what to add to the negative prompts to lessen
I think i got to the bottom of this mostly I had a model suggest to add this to the positive prompt "photography shot with Canon EOS 5D Mark IV". I'm guessing there is a big high quailty data set with this tag and it is in most of my checkpoints. it would add a camera to the scene and this Lora would morph it into some weird accessories.
I got rid of this mostly(95%) by first adding parenthesis around (photography shot with Canon EOS 5D Mark IV) like so to try and limit it from breaking it into multiple tokens, then i added all of these to the negative prompt: aviator sunglasses , aviator cap, aviator goggles, leather armor, leather mask, leather, leather strap, one piece, stand \\(jojo\\), jojo no kimyou na bouken, dio brando, speech bubble, thought bubble,robots,gadgets,deep wound, wound, blood, scar, stitches, surgical,(Canon EOS 5D Mark IV)
hope it helps others
Looks great but I can't use it, seems like the download is corrupted. Keeps telling me the header is too small.
SOLVED: I'm stupid, I had it in both the checkpoint AND the Loras folder for comfyui. Deleted it from checkpoint, worked like a charm. I'll leave this in case anyone else has the same issue at some point.
This looks great, but it completely smears eyes, teeth, and sometimes lips if the character's face in not very close to the viewer. Any suggestions how to go around this?
are you using adetailer?
@durachell No
@penguin233 https://github.com/Bing-su/adetailer
@durachell ok, I see, this helps massively, thanks.
great stuff, this improves any sdxl model apparently
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.