All illustrations are free for commercial use. Please feel free to use them for announcements or as eye-catching images. ↑ ↑ ↑
【Stable Diffusion】Great Compatibility with LoRA! Top 3 Recommended New Models
Hello, this is kojirom.
Today, I will introduce three carefully selected models from the newly released ones.
This is a 2.5D style model. According to the creator’s page, it is said to be ‘compatible with LoRA,’ and upon testing, I found this to be true. In fact, in this blog, we download over 10 models each time and generate images with them. During this process, we try various applications of LoRA and exclude models that break down severely. So, you can trust the models introduced here.
This model is highly stable, so I expect it will become quite popular. For example, when you input ‘wearing a swimwear’ in the prompt, many models generate black or navy swimwear, but this model produced swimsuits in a variety of shapes and colors.
It creates well-crafted illustrations, impressive for a debut work. You can enjoy characters drawn with delicate lines and elaborate backgrounds. Let’s look forward to future updates.
Please find your favorite models and illustrations on our blog.
- “The images in this article were generated without the use of any face correction features.”
- “The prompts used were only ‘a single girl’ and ‘in a swimsuit’.”
- “For negative prompts, we utilized the ‘EasyNegative2’ embeddings.”
- “On our blog, we only feature models that allow commercial use of their generated images.”
- “Please feel free to use the illustrations within the bounds of common sense and good judgement.”
“‘Blazarot’ is a model that combines BlazingDrive10g with Kakarot2.5DCozy. I liked the semi-realistic facial style of Kakarot and merged it with another model to improve the use of color. This model focuses on the appearance of semi-realistic characters and is compatible with Danbooru style tags. Due to Kakarot’s influence, the facial tones can be a bit pale, but this has been adjusted in later versions. It is suitable for anime-style illustrations that seek a ‘realistic’ body and semi-realistic face.”
“MW_Anime Painting Anime 1.1 is an anime model that strongly recommends the use of high-resolution restoration features. For R-ESRGAN 4x+Anime6B, a redraw amplitude of 0.5 is recommended, and the CLIP Layer is set to 2. Euler a and DPM++2M Karras are suggested as sampling methods, and you can also try other methods. If you like this model, feel free to give it a like and share your beautiful images. You can also share your experiences with AI painting and SD in the QQ group.”
note: Site Summary
“We are working on creating a model to help artists work faster. It aims for a design with many lines and a sense of space, but it is still in the beta stage. Please feel free to try it out.”