Instruction:
Recently, I was delighted to discover a highly effective model for anime image generation on the website—the illustrious model. This model produces images with remarkable detail and quality. For the following content, I will be using this model to demonstrate its capabilities.
About the “illustrious Model”:
The illustrious model is an advanced AI model specifically designed for creating illustrations and anime-style images. Developed by the OnomaAI research team, it is based on the Stable Diffusion XL (SDXL) architecture and has been optimized to enhance the quality of illustration and animation tasks. Its key features include high-resolution image generation, multi-level prompt support, and an open-source nature.
Thanks to its development using the large-scale Danbooru2023 dataset—which encompasses a vast range of character designs and artistic styles—there is often no need to include additional Loras to generate well-known and distinguished characters.
However, using Loras can be beneficial for accurately depicting specific outfits, maintaining a consistent style, and making it easier to add or remove certain elements or details in the attire.
In this article, I will use the illustrious model to demonstrate image generation and explore how Lora influences character clothing elements and details.
Evaluate from these three perspectives：
- Without using a Lora: You need to explicitly mention the character's name in the prompt for the model to generate the desired character.
- Using a Lora: When triggering a Lora (if it requires a trigger word), you do not need to mention the character’s name.
- Using a Lora and mentioning the character: When triggering a Lora (if it requires a trigger word) and mentioning the character in the prompt.
Each of these methods can achieve effective results, and the choice ultimately depends on your preference. However, keep in mind that mentioning the character’s name while using a Lora can significantly impact the outcome.
Base example：
1girl, blonde hair, long hair, blue eyes, forehead, sidelocks, braid, keyhole, forehead mark, polka dot bow, hair bow, parted bang, grey dress, long sleeves, skirt, bodystocking, gloves, beret, bare legs, holding lantern, serious, night, hallway, mansion, night sky, window, indoors, cowboy shot, blushing
Prompt 1：
Add Lora without mentioning the character.
1girl, blonde hair, long hair, blue eyes, forehead, sidelocks, braid, keyhole, forehead mark, polka dot bow, hair bow, parted bang, grey dress, long sleeves, skirt, bodystocking, gloves, beret, bare legs, holding lantern, serious, night, hallway, mansion, night sky, window, indoors, cowboy shot, blushing, <lora:Abby_Illus:1>
prompt 2:
Mention the character without adding Lora.
1girl, blonde hair, long hair, blue eyes, forehead, sidelocks, braid, keyhole, forehead mark, polka dot bow, hair bow, parted bang, grey dress, long sleeves, skirt, bodystocking, gloves, beret, bare legs, holding lantern, serious, night, hallway, mansion, night sky, window, indoors, cowboy shot, blushing, abigail williams \(fate\)
prompt 3:
Add the character and their traits.
1girl, blonde hair, long hair, blue eyes, forehead, sidelocks, braid, keyhole, forehead mark, polka dot bow, hair bow, parted bang, grey dress, long sleeves, skirt, bodystocking, gloves, beret, bare legs, holding lantern, serious, night, hallway, mansion, night sky, window, indoors, cowboy shot, blushing, abigail williams \(fate\), abigail williams \(event portrait\) \(fate\)
prompt 5:
Add the character, character traits, and Lora.
1girl, blonde hair, long hair, blue eyes, forehead, sidelocks, braid, keyhole, forehead mark, polka dot bow, hair bow, parted bang, grey dress, long sleeves, skirt, bodystocking, gloves, beret, bare legs, holding lantern, serious, night, hallway, mansion, night sky, window, indoors, cowboy shot, blushing, abigail williams \(fate\), abigail williams \(event portrait\) \(fate\) <lora:Abby_Illus:1>
As you can see, to generate specific clothing more accurately and achieve better image results, it’s best to use Lora without mentioning the character’s name. If you only mention the character without using Lora, the generated clothing might not match your desired outcome. On the other hand, using Lora alone ensures that the clothing details are rendered with greater precision and perfection.
I must say, this model is highly practical, producing images of exceptional quality. I encourage everyone to try using this method to modify character clothing or styles and experience the impressive results this model offers!
https://www.seaart.ai/models/detail/csl2e7le878c73dgjin0
