Animagine XL is a high-quality model focusing on anime-style images. The model has deep knowledge of various characters from well-known series. It excels in creating accurate and detailed characters to serve anime fans and AI artists, with its unique and easy tag system, users can easily create amazing images.
What This Model Excels At
- Generating anime style images.
- Generating exist anime characters with accuracy and details.
- Generating female/male images.
- Crafting SFW/NSFW images.
- Crafting impressive landscapes.
- Working good with flexible prompt order.
- Has special tags for better prompting experience.
- Supports Danbooru based tags.
- Works better at certain resolutions.
Cautions For Optimal Usage
- Requires VAE for better generations.
- Short and simple prompts can cause wrong or low quality results.
- Requires long detailed prompts to generate desired images.
- Use Special Tags for easier and better results, but model can also generate without them.
- Prefer Danbooru tags instead of general syntax, since model optimized for them.
- Can create NSFW content without prompting it.
- Improved but still can struggle with anatomy and hand generation in some images.
- Far distance shots can reduce quality of subject.
Recommended Settings
- Steps: Under 30 (mostly 28)
- CFG Scale: 5-7
- Sampler: Euler a
- Clip Skip: 2
- VAE: sdxl_vae, PyTorch_vae (not available yet)
Supported Resolution And Aspect Ratio
- `1024x1024`-`[1:1]`
- `1152x896`-`[9:7] H`
- `896x1152`-`[7:9] V`
- `1216x832`-`[19:13] H`
- `832x1216`-`[13:19] V`
- `1344x768`-`[7:4] H`
- `768x1344`-`[4:7] V`
- `1536x640`-`[12:5] H`
- `640x1536`-`[5:12] V`
Suggested Resources
LoRA
- Users can use LoRAs, if they are not able to get desired result with models training.
Prompt Style
- Use long and descriptive prompts.
- Use special tags to achieve high quality results.
- Use Danbooru tags for better results.
- Prompt recommendations for high quality, aesthetic images:
Positive Prompt
`masterpiece, best quality, very aesthetic, absurdres`
Negative Prompt
`nsfw, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract]`
Recommended Prompt Order
- Subject (1girl/1boy-landscape/scenery)
- Character Name
- From Which Anime/Game
- Everything else in any order
- Quality Modifiers
Special Tags
Animagine XL uses special tags to guide the result towards quality, rating, creation date and aesthetics. The model can generate images without these tags, but using them can help achieve better results.
Quality Modifiers
Quality tags consider both scores and post ratings to ensure a balanced quality distribution.
- 100% ≥ `masterpiece` > 95%
- 95% ≥ `best quality` > 85%
- 85% ≥ `great quality` > 75%
- 75% ≥ `good quality` > 50%
- 50% ≥ `normal quality` > 25%
- 25% ≥ `low quality` > 10%
- 10% ≥ `worst quality` > 0%
Rating Modifiers
Rating tags control safety of content.
- `safe`: General
- `sensitive`: Sensitive
- `nsfw`: Questionable
- `explicit, nsfw`: Explicit
Year Modifiers
Year tags are used to more accurately direct results to specific modern or classic anime art styles.
- `newest` from 2021 to 2024
- `recent` from 2018 to 2020
- `mid` from 2015 to 2017
- `early` from 2011 to 2014
- `oldest` from 2005 to 2010
Aesthetic Tags
Aesthetic tags can be used to refine generations. Scores based on visual appeal.
- 1 ≥ `very aesthetic` > 0.71
- 0.71 ≥ `aesthetic` > 0.45
- 0.45 ≥ `displeasing` > 0.27
- 0.27 ≥ `very displeasing` > 0
Prompt Examples
Esdeath
Positive Prompt
`1girl, esdeath, akame ga kill!, :d, blue eyes, blush, blue hair, long hair, chest tattoo, white uniform, military hat, cleavage, claw pose, looking at viewer, sky scenery, clouds, sunshine, horizon, dramatic lighting, soft shadow, masterpiece, best quality, safe, SFW, very aesthetic, recent, absurdres, highres`
Negative Prompt
`(worst quality, low quality:1.3), NSFW, explicit, oldest, early, (very displeasing, displeasing), lowres, bad anatomy, anatomical nonsense, bad hands, jpeg artifacts, ((ugly)), grayscale, traditional media`
Guts
Positive Prompt
`1boy, guts, berserk, close-up, evil smile, (one eye closed:1.2), scar on face, short hair, black hair, greyscale, glowing eye, red eye, dark aura, ((by kentaro miura, manga)), dark ambient, amazing background, masterpiece, best quality, very aesthetic, newest`
Negative Prompt
`(worst quality, low quality:1.3), NSFW, explicit, oldest, early, (very displeasing, displeasing), lowres, bad anatomy, anatomical nonsense, jpeg artifacts, ((ugly))`
You can join SeaArt Discord Community for more content, also you will have opportunity to choose next model guides with your vote. You can use comments or SeaArt server #guides-feedback to leave feedback or suggestion. We look forward to seeing you on the server.
Happy Generations To Everyone!