How to Create Anime Art with AI in Problembo
Create your first anime image
Open the AI anime generator, then follow these steps:
- Sign in. You can inspect the catalog and prices without an account, but an account is required to start a generation.
- Describe the character or scene. Include appearance, clothing, action, environment, light, and camera angle.
- Choose a model and its available settings. The form changes when a model supports styles, reference images, or a negative prompt.
- Choose an aspect ratio and the number of variants.
- Check the exact price shown before launch, run the task, and download the results you want to keep.
Each run is charged separately; there is no subscription. Supported models can use the daily bonus balance after sign-in.
Choose an anime model
| Model | Start with it when you need | Available controls |
|---|---|---|
| WaifuStudio v2.6 | Detailed anime characters and stylized scenes | Styles, reference, negative prompt |
| Anime XL v9 | Detailed characters and expressive anime art | Reference, negative prompt |
| Animake v1.1 | Fast character, illustration, and scene drafts | Styles, up to 8 variants |
| Velvet Render v5.1 Anime | Detailed anime-style illustrations | Negative prompt, up to 4 variants |
| Manga Master v4 | Manga, anime art, and furry-style images | Negative prompt, up to 4 variants |
| Rev Animated v2.7 | Anime, stylized, and more realistic illustrations | Reference, negative prompt |
WaifuStudio v2.6, Anime XL v9, and Rev Animated v2.7 can use a source image as a reference. Animake v1.1 accepts up to 8 variants; WaifuStudio v2.6, Anime XL v9, and Rev Animated v2.7 accept up to 12. The form is the source of truth because the catalog and model availability can change.
Two models have dedicated instructions:
Build the prompt in a stable order
A prompt is easier to revise when every part has a purpose:
[character count], [appearance], [clothing], [pose or action],
[environment], [lighting and palette], [camera and composition]
Example:
1girl, solo, short silver hair, green eyes, dark blue coat, holding a paper lantern,
snowy city, evening, soft warm light, full body. She is walking toward the viewer as snow falls.
Put identity-defining details near the beginning. Use one main action and one framing. If you ask
for close-up, full body, and wide shot together, the instructions compete with each other.
Use tags where the model supports them
Animake v1.1 works with English Danbooru tags, natural-language descriptions, or both. Write tags in lowercase and use spaces instead of underscores. A useful order is:
[quality/year/safety], [characters], [appearance], [clothing], [action], [environment]
Useful examples include:
1girl,1boy,solo— character count and groupingfull body,portrait,looking at viewer— framing and directionyear 2025,newest— a contemporary visual treatment in Animake v1.1safe,sensitive,nsfw,explicit— intended content rating in Animake v1.1
For Manga Master v4, tags such as high complexity, depthness, year 2025, and
transparent background have model-specific meanings documented in its guide. Do not assume that a
tag will behave the same way in every model; replace ineffective tags with a plain description of
the visual feature you need.
Write a focused negative prompt
WaifuStudio v2.6, Anime XL v9, Velvet Render v5.1 Anime, Manga Master v4, and Rev Animated v2.7 expose a negative prompt. Start with visible problems you want to avoid:
text, watermark, logo, extra fingers, bad hands
Keep it short and do not ban something requested in the main prompt. For a vivid drawn look in Manga Master v4, its guide suggests:
realistic, photorealistic, 3d, muted colors
A negative prompt can reduce unwanted details, but it does not guarantee a flawless result.
Choose an aspect ratio for the composition
The anime catalog offers five aspect ratios:
| Ratio | Typical composition |
|---|---|
| 1:1 | Square avatar, face, or centered character |
| 3:4 | Portrait or full-body character |
| 4:3 | Character with a wider environment |
| 9:16 | Tall poster or mobile composition |
| 16:9 | Wide scene, group, or landscape |
Choose the ratio before refining composition words. If the result feels cramped, move from a square to a portrait or landscape frame rather than adding conflicting camera instructions.
Keep a character consistent
AI generation does not guarantee an identical character across separate runs, but a stable prompt reduces variation:
- Write a short character sheet: hair, eyes, face, build, clothing, accessories, and core colors.
- Copy those traits unchanged into every prompt.
- Keep the same model. Models interpret the same words differently.
- Keep framing, palette, and lighting stable for images in one series.
- With WaifuStudio v2.6, Anime XL v9, or Rev Animated v2.7, reuse the same clear reference and state whether it controls the face, outfit, pose, or palette.
- Change one variable at a time and generate a small batch before expanding the scene.
Reference-enabled models accept up to 4 JPG, JPEG, PNG, WebP, AVIF, HEIC, or HEIF files, up to 15 MB each. A reference guides the model but does not copy every detail exactly.
Common mistakes
- A prompt that is too vague. Add fixed visual traits, an action, a setting, and lighting.
- Contradictory framing. Choose one of portrait, close-up, full body, or wide shot.
- Too many changes at once. Revise one part of the prompt so you can see what affected the result.
- Relying only on a character or artist name. Describe visible traits and the intended result.
- Using unsupported controls. If the form does not show a reference, style, or negative prompt, that model does not expose it.
- Expecting long readable text. Animake v1.1 may render individual words but is not intended for long text.
- Forgetting to download results. Files are normally stored for 3 days, or for 30 days if the paid balance is at least $1 when they are stored.
When the prompt is ready, return to the AI anime generator, compare the model examples, and check the quote before starting the run.