chenkin-noob-xl-prompting

name	chenkin-noob-xl-prompting
disable-model-invocation	true
description	Prompting rules for **Chenkin Noob XL (CKXL)**, a NoobAI-XL-1.1 *eps* fine-tune. Use when the user asks to generate / draw / render an anime image on a CKXL workflow (v0.2, v0.5, or near-vanilla finetune). CKXL is eps (not v-pred) and uses `artist:name` syntax with anti-furry negatives; it adds CKXL-specific quality tags (`aesthetic`, `excellent`).

Chenkin Noob XL Prompting

Chenkin Noob XL (CKXL) is a Chenkin fine-tune of Laxhar/noobai-XL-1.1 (an eps-prediction SDXL anime model), trained on ~12M images (9M anime + 2.17M game/Western designs, Danbooru through Jan 2026). It uses a Danbooru + e621 tag namespace with artist:name syntax, and adds a CKXL-specific aesthetic/quality vocabulary (Aesthetic, Excellent, High Resolution, Medium Resolution). Write the prompt and negative you pass to the generate_image tool per the rules below to drive it well on the first call.

CKXL is eps-prediction, not v-pred - it wants the eps sampler / CFG settings below (Euler a, CFG 5-6, no zsnr), not v-pred constraints.

Safe default when unsure: use the CKXL workflow; lead the prompt with masterpiece, best quality, newest, high resolution, aesthetic, excellent, year 2026,, add roughly 8-15 caption-order tags for the subject, and always pass the recommended negative below including its anti-furry block. CKXL has no aggressive default style, so terse prompts come out bland - spend the detail. Everything below is how to do better than this default when the request calls for it.

The generate_image call

Arg	For CKXL
prompt	Required. Lead with the CKXL prefix below; then the caption order.
negative	Required. Start from the recommended negative below - keep the anti-furry block.
workflow	Use the CKXL workflow (or an eps-prediction workflow with the CKXL checkpoint pointed at by it).
width / height	Total area ~1024x1024. Best: 768x1344, 832x1216, 896x1152, 1024x1024, 1152x896, 1216x832, 1344x768.
steps	25-30.
cfg	5-6. Higher oversaturates; stay <= 6.
seed	Omit for a fresh random image; pass a prior seed to reproduce or vary.

Do not pass inputImage unless the workflow is a CKXL img2img variant.

Positive prompt recipe

1. Required prefix (v0.5 recommendation): masterpiece, best quality, newest, high resolution, aesthetic, excellent, year 2026,
   • For older-looking output, swap year 2026 for an earlier year (year 2023) or a period tag (see Date tags).
   • aesthetic, excellent, high resolution are CKXL-specific quality tags - keep them. Drop them only if you want a deliberately less-polished look.
2. Caption order: <1girl/1boy/1other/...>, <character>, <series>, <artists>, <special tags>, <general tags>, <other tags>
3. Artists use artist:name syntax, comma-joined: artist:wlop, artist:as109. Multi-word artist names keep underscores (artist:john_kafka). Do not use an @ prefix.
4. Lowercase tags, spaces over underscores in general tags; multi-word Danbooru proper nouns may keep underscores.
5. Escape literal parentheses with backslash: arlecchino \(genshin impact\).
6. Name then describe characters. With multiple characters, describe each separately to avoid conflation.
7. Be specific. CKXL has no aggressive default style, so terse prompts come out bland.

Negative prompt recipe

Start from the model card's v0.5 recommendation and add the anti-furry block:

nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands,
mammal, anthro, furry, ambiguous form, feral, semi-anthro, e621

The v0.5 card itself drops the mammal, anthro, furry... block, but CKXL is built on a checkpoint trained on e621, so keep it unless the user actually wants furry/anthro art. The v0.2 card's e621, Furry shorthand goes in the same role.

Add situational terms:

• text, watermark, multiple views, extra digits, jpeg artifacts for cleanup.
• Drop nsfw if the user wants sensitive/explicit output.

Tag reference

CKXL quality vocabulary

Tag	Meaning
masterpiece	Top 5% quality percentile - keep in the prefix
best quality	Top 15% quality percentile - keep in the prefix
aesthetic	CKXL top-100k Danbooru aesthetic - keep in the prefix
excellent	CKXL top-1M liked Danbooru - keep in the prefix
high resolution	CKXL >= 2048 px source - default for the prefix
medium resolution	CKXL 1024-2048 px source - lower-resolution training look

Date tags

Period tag	Year range	When to use
newest	2022-2025	CKXL's recommended modern look - the default
recent	2018-2020	Late-2010s / early-2020s style
mid	2014-2017	Mid-2010s style
early	2011-2014	Earlier 2010s style
old	2005-2010	Vintage 2000s anime

For a specific year, use year 2024, year 2025, year 2026 etc. CKXL's training data runs through Jan 2026, so year 2026 is in-range.

Generation settings

Setting	Value (per the model card)
Resolution	~1024x1024 area. Best: 832x1216 portrait, 1024x1024 square, 1216x832 landscape.
Steps	25-30.
CFG	5-6. Higher oversaturates.
Sampler	Euler a (the eps-line default - not plain Euler, which is the v-pred constraint).

Sampler / scheduler live in the workflow file. CKXL is eps-prediction - if the workflow is configured for v-pred (zsnr, v_prediction), generations will look mushy. Verify the workflow before retrying.

Prompt enhancement (enhance)

The generate_image tool has an opt-in enhance option that routes your prompt through a separate model before the render to rewrite it into CKXL's native protocol - the prefixed, caption-ordered Danbooru / e621 tag list this guide describes.

• It is opt-in. Pass enhance: true on the call (or a workflow/config default turns it on). When off, your prompt is sent as-is.
• Use it when the incoming prompt is thin (a few words) and you want a one-shot upgrade to a full tagged prompt. When you have already built a careful tagged prompt, leave enhance off - it only adds latency and risks drifting from your intent.
• Scene continuity. Pass the context arg alongside enhance to hand the enhancer background to honour (character facts, ongoing scene, wardrobe) without depicting it literally. It is ignored when enhance is off.
• Negative. The enhancer builds on whatever baseline negative you pass and returns a refined one - it must keep the anti-furry block (see the negative recipe).

If you are enhancing a prompt rather than rendering

You may be reading this not to call generate_image yourself, but as guidance handed to the prompt-enhancement step: you were given a rough positive prompt, a baseline negative, and the target workflow's protocol, and you must return a single JSON object {"prompt", "negative"} and nothing else. In that role:

• Act only on the prompt-writing rules above - the required prefix and caption order in the positive recipe and the negative recipe. The generation-settings and tool-arg tables do not apply - you do not pick a workflow or call any tool.
• Translate and enrich, do not reinvent. Keep the incoming subject and intent; lead with masterpiece, best quality, newest, high resolution, aesthetic, excellent, year 2026, (keep the CKXL quality tags), write artists as comma-joined artist:name (no @), use lowercase tags, and escape literal parentheses (\( \)).
• Always add detail - even to an already-tagged prompt. CKXL has no aggressive default style, so push toward roughly 8-15 caption-order items: subject count (1girl / 1boy / 1other), character / series, appearance, pose, scene, and lighting that the prompt leaves unstated, staying faithful to the stated subject.
• Mine any background context you were given (a scene/continuity note, recent conversation) for that extra detail and fold it in as depictable tags. Treat it as source material to pick from, not a checklist to dump and not a subject that overrides the explicit prompt; ignore chatter with no visual bearing.
• Build on the baseline negative and always keep the anti-furry block (mammal, anthro, furry, ambiguous form, feral, semi-anthro, e621) unless the user actually wants furry / anthro art.

Worked examples

These illustrate the shape of a finished prompt - adapt them to the actual request, do not reuse them verbatim. Pass the prompt (and negative) value as a single string.

Standard portrait:

prompt:   masterpiece, best quality, newest, high resolution, aesthetic, excellent, year 2026,
          1girl, solo, fern \(sousou no frieren\), sousou no frieren, artist:wlop,
          long purple hair, purple eyes, black coat, holding staff, magical sparks,
          forest clearing at dusk, cinematic lighting, soft glow, looking at viewer
negative: nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands,
          mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro, e621,
          text, watermark

Multi-artist style mix:

prompt:   masterpiece, best quality, newest, high resolution, aesthetic, excellent, year 2025,
          artist:wlop, artist:as109,
          1girl, solo, chromatic aberration, film grain, horror \(theme\), limited palette,
          x-shaped pupils, high contrast, cold colors, black theme, gritty
negative: nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands,
          mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro, e621

The wrapped lines above are a single comma/space-joined string each - pass them as one line in the tool args.

Anti-patterns

• Treating CKXL like a v-pred model. CFG 4 + Euler + zsnr underperforms here - CKXL is eps and wants CFG 5-6 + Euler a + no zsnr.
• Dropping CKXL's quality tags. aesthetic, excellent, high resolution are this fine-tune's main aesthetic lever - removing them defeats the point of this fine-tune.
• Skipping the anti-furry block. Even though v0.5's card omits it, CKXL is built on an e621-trained checkpoint - without the block, e621 features leak into anime prompts.
• @-prefixed artists. Wrong here - use artist:name instead.
• Pushing CFG past 6. Oversaturates the image on the eps line; CKXL is no exception.
• Three-word prompts. CKXL has no aggressive default style, so terse prompts come out bland. Always include the full prefix + caption-order tags.
• Wrong year tag. year 2027 is out of CKXL's training range (training data cutoff is Jan 2026); the tag will be treated as unknown.
• Square-bracket weighting. ComfyUI weights with parentheses - (tag:1.3) to strengthen, (tag:0.8) to weaken. A1111-style square brackets do not de-emphasize here: [tag] parses as ([tag]:1) (literal brackets at weight 1). Use (tag:0.8) to weaken instead.