The long overdue V2.0 Turbo version is here!
This time still optimized for detail and generalizability, thanks to the Turbo model's ability to greatly reduce the greasy feeling of lora, come and try it out! Can you give me a like if you can, please?
Recommended Parameters:
Sampler: Euler
Scheduler: SGM_Uniform
CFG: 1
Image Size: 1024x1536
Number of Samples: 25
Model Weight: Between 0.6 and 0.8
I heard that FLUX came out with a Krea version, and immediately trained a Lora to see the results! The parameters of the generated images are consistent with the Dev version!
This time for a taste!
This time, I took the opportunity of retraining to enhance the details of the 2.0 version!
The new V3.1 version of Semi-Realistic Portraits is here!
1. Add more details to the picture
2. Further reduce the greasiness of the skin
3. Further reduce the sense of redness on the face
4. Add new Chinese elements and scene support, utilize your unlimited creativity!
V3.0 version
This update is a bold attempt to add more artistic creativity! It is also a brand new change!
1, I combined with everyone's most used portraits, made enhancements for young women, but accordingly, the support for other genders as well as age groups may not be as strong as in the past.
2、This time, many artistic elements have been added, which can be illustrations, realistic photos, or a combination of illustration and reality, depending on your description!
3、Some corrections have been made for hand problems
4、Generalization is enhanced! You can use it with different models, there will be unexpected effects!
5. Future direction: 1. reduce the redness of the face, 2. add more details to the face, 3. continue to optimize the hand and foot blemishes, 4. reduce the greasiness.
Finally, it's not easy to create, I hope that if you like it, you can point a collection or praise, you can also give me back to the picture, your support is my motivation to update!
The semi-realistic colored pencil portrait Lora V2.0 version is here!
New portraits of people at different stages have been added: children, kids, young people, middle-aged people, and elderly people, covering different stages of human growth.
Semi-realistic colored pencil portrait LoRA is here!
This LoRA has a relatively unrestrained and relaxed painting style. Only the face is painted in a realistic way, while other parts of the face and clothes are in a semi-realistic way, fully highlighting the facial beauty.
It is mainly for portraits of Europeans and Americans. If you need portraits of people from different regions, you need to add corresponding keywords: 1. Asian face, 2. African face, 3. Australian face, and so on. It can be combined with other realistic or facial models, which will produce unexpected effects. Looking forward to everyone's download and use.
Description
The semi-realistic colored pencil portrait Lora V2.0 version is here!
New portraits of people at different stages have been added: children, kids, young people, middle-aged people, and elderly people, covering different stages of human growth.
FAQ
Comments (14)
This is a great model. I like the colors and feel that it gives off. Thank you!
The not so good is that it's a bit bad on landscape mode. If you check the posts I made, you'll see why. The result can be good but it's a low chance, I either got split images or different colors or just plain old white background.
I also noticed that sometimes it doesn't follow the prompt if the prompt is on the longer side. Haven't had this problem with other Loras and the checkpoint I am using.
For reference, I am using flux1-dev-Q8_0.gguf checkpoint.
I tried a couple more landscape and they came out great and now I don't know what to think anymore ;)))
I'll upload them in a bit.
Thanks for your input, this lora is more inclined to portraits, maybe because the dataset is not broad enough yet, I will try to fix him in version 3.0
@1jiliguala1 I uploaded some new landscape ones, which are better, I only used this Lora for the others it was a combination of Loras and Embeddings. Maybe because it's Flux it doesn't do that well with negative prompts. For these ones I didn't use any negative prompts and they came out great! I like how the change of contrast affects the whole mood of the image.
This Lora generates mesmerizing semi-realistic portraits of humans and humanoids (anthro).
Thanks for the like!
do you have a pony version?
Hello. Not yet.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.