AI model

NoobAI XL Epsilon 1.1: familiar epsilon guidance

Choose this when you want a Noob look with standard SDXL-style CFG habits: portraits, city nights, and bright daylight scenes with modern anime sheen.

100 free credits High resolution Commercial rights

Names are descriptive only, not a license, sponsorship, endorsement, or affiliation. We are not the publisher of third-party model weights, except files we host. You are responsible for your use, per the model's license and our terms. As-is, no warranty of non-infringement (not legal advice). For DMCA, copyright, or trademark issues, use our Contact page.

Developer
Community / Civitai distribution
Release date
November 20, 2024
Base model
Stable Diffusion XL 1.0
Training data
NoobAI XL training lineage (epsilon profile)

About NoobAI XL (Epsilon 1.1)

NoobAI XL (Epsilon 1.1) is a Civitai-distributed Noob family checkpoint with epsilon-style prediction. It shares the “masterpiece, best quality, newest, absurdres, highres” stack with other Noob entries but trades the V-pred specific guidance range for a more conventional CFG curve, handy for artists migrating from other epsilon SDXL models.

NoobAI XL Epsilon 1.1 is the epsilon-prediction sibling in the Civitai-distributed Noob family. It shares the modern-anime DNA and the familiar "masterpiece, best quality, newest, absurdres, highres" vocabulary with the rest of the line, but uses standard SDXL-style guidance behaviour rather than the V-prediction curve. For artists migrating from other epsilon SDXL anime checkpoints, that means the parameters you already trust mostly transfer over: a CFG in the mid-range, around thirty steps on Euler A, and the same negative prompt habits you developed elsewhere.

Compared to the V-prediction Noob build, Epsilon 1.1 tends to be more tolerant of mid and higher guidance scales, and can produce a touch sharper micro-contrast on faces and clothing folds. It is a strong pick for cinematic night scenes, neon-lit portraits, train station settings, half-body shots with shallow depth of field, and other prompts where you want decisive edges rather than the smoother large-gradient look of V-prediction.

If you are choosing between the V-pred and Epsilon Noob builds, treat them as guidance and contrast options on the same family aesthetic. Lock the seed and resolution, run the same prompt through both, and pick the one whose contrast and skin tone match the post-production look you want for the project.

Example generations

See what NoobAI XL (Epsilon 1.1) can create with the right prompts

NoobAI XL (Epsilon 1.1) example 1
1
832x1216

NoobAI XL (Epsilon 1.1), masterpiece, best quality, newest, absurdres, highres, 1girl, detailed eyes, half body, warm lighting, bokeh

Try this prompt
NoobAI XL (Epsilon 1.1) example 2
2
832x1216

NoobAI XL (Epsilon 1.1), masterpiece, best quality, newest, absurdres, highres, 1girl, twin braids, sweater, coffee shop, cozy

Try this prompt
NoobAI XL (Epsilon 1.1) example 3
3
832x1216

NoobAI XL (Epsilon 1.1), masterpiece, best quality, newest, absurdres, highres, 1boy, sharp eyes, suit, office window, cityscape

Try this prompt
NoobAI XL (Epsilon 1.1) example 4
4
1216x832

NoobAI XL (Epsilon 1.1), masterpiece, best quality, newest, absurdres, highres, train station at dusk, no people, cinematic, rain wet ground

Try this prompt

Key features

⚙️

Epsilon prediction

Predicts in the epsilon / noise parameterization style; pairs with the app’s non-vpred FAL path.

🔦

Noob “newest / absurdres”

Same family vocabulary: newest + absurdres + highres to anchor a contemporary anime “HD” look.

📐

SDXL layout habits

Stack subject, then wardrobe, then environment and lighting. The checkpoint rewards a clear order of attention, especially for portraits, bust shots, and city street scenes.

🌃

Cinematic night scenes and bokeh

Strong match for city lights, train stations, and half-body portraits with shallow depth. Good anchor keywords: neon bokeh, wet ground, studio light.

Faster iteration without V-pred

Choose this epsilon build when you want Noob color habits but a parameter window closer to other SDXL anime checkpoints on the same scheduler.

🧾

Consistent quality block

Reuse the same "newest, absurdres, highres" stack across prompts so a batch of variants lands on a similar quality bar instead of drifting between runs.

Prompting guide

Learn how to get the best results from NoobAI XL (Epsilon 1.1)

1

Start near CFG 7 with Euler A

The model defaults in Anifusion target ~28 steps. Adjust CFG in small steps if edges feel harsh.

2

Use the classic quality tag block

Pair your scene tags with the usual quality stack, such as masterpiece, best quality, newest, absurdres, and highres, so the run stays on-model.

Recommended settings

Scheduler
Euler A
Steps
28
Guidance scale
7
Image sizes
1024x1024, 832x1216, 1216x832
Quality tags (add to prompt)
masterpiece, best quality, newest, absurdres, highres
Negative prompt
lowres, bad anatomy, blurry, worst quality, low quality, bad hands, mutated hands, extra arms, duplicate, cropped, watermark, signature, ugly, monochrome

Recommended tags

masterpiecebest qualitynewestabsurdreshighres1girl1boybokehcinematicdetailed background

Frequently asked questions

Which Noob do I pick on Anifusion?

Use Epsilon 1.1 for epsilon-habit workflows, Civit V-pred for the V-pred profile, and the base mirror for the default hosted weight. Compare outputs, not the label alone.

Where should I start with guidance and steps?

Start near guidance 7 with Euler A and about 28 steps. Nudge guidance in small steps if edges feel harsh, before you raise step count or switch samplers.

Start creating with NoobAI XL (Epsilon 1.1)

Join 130,000+ creators and bring your anime art to life with 100 free credits.

130,000+ creators 10M+ images generated Full commercial rights