3 ways to generate more beautiful anime-style illustrations with Image V2

As I write this article, in the community, SNoW's post discusses the gallery posting function and model usage restrictions, including proposals to charge for using certain models or for posting itself...

If it ever happens that free access to each model becomes limited, we may need to generate higher-quality images with fewer opportunities 😓

Therefore, as someone who has long used the image V2 (hereafter V2) model, which has been on this site since its early days and seems likely to remain, I will write down tips for generating beautiful anime-style illustrations with minimal defects using V2 for future users' reference 🖋️

① Use prompts that describe light ✨

As I wrote in , V2 excels at depicting light. Therefore, it is recommended to write prompts related to light depiction, such as "sparkling," "lens flare," or "rim light"! You can also supplement them with the auto-prompt function!

↓ Example 1-1: Anime CG, cute girl, long hair, smile, white long-sleeved clothes, daytime garden, fantastical light, sunlight filtering through trees, sparkling, lens flare, rim light

↓ Example 1-2: Anime CG, dark fantasy, girl, white dress with delicate lace, night garden, deep darkness, fantastical pale light, moonlight, sparkling, lens flare, rim light

② Create a sense of perspective 📷️

V2 is good at depicting blurring of distant or close objects! So for portraits, adding a prompt describing the background scenery can create a more profound image!

↓ Example 2: Anime CG, cute girl, smile, sitting in a field, kimono, spring flowers, distant hazy mountains, blue sky, lens flare

③ My personal favorite 6-element prompt 🔮

Anime CG, girl, inside a dream, field, flowers and grass, sparkling

↓ Example 3

Despite being a prompt consisting of just six elements, this alone can bring out V2's strengths! Hope it helps ✨
(Also, based on experience, V2 tends to output prettier images with shorter prompts.)

【Conclusion】

Whether the above methods work for other models depends on the model, but I hope it can be of some help for everyone!
Also, prompts from the heyday of V2 last year or the year before are a great reference 😆

【Bonus】

One long-standing issue with V2 when generating anime-style illustrations has been occasional generation defects like finger bugs, but I have been gradually finding ways to improve this ⚙️
For example, using Anime CG instead of "anime style" seems to reduce bugs. I plan to compile this into another article in the future 🖊️

↓ Example 4: Anime CG, cute girl in a black school uniform standing in the hallway of a wooden school building, lens flare