Zhenzhen's AI Radar · MOSS Image · 2026-11-13 Morning Report
This issue features 43 new versions: Strongly Recommended 1 · Recommended 8 · Worth Watching 16 · Included Only 18
Editor's Note This issue reviews 47 new MOSS versions, with 1 rated S, 9 rated A, 17 rated B, and 19 rated C, excluding 1 duplicate uploads. Top pick: BAStyle v2.0: A clean cel-shaded style in the vein of Blue Archive. After re-annotating the training images, it corrects the outdated facial structures and oily hair textures seen in single-character generations. Strength, steps, CFG, and workflows are all provided. DecoSplosion collage decoration style, Deadpan minimalist chibi doodles, V67 watercolor, 12-step Turbo v9, and light/shadow sliders are rated A. Several character LoRAs automatically rated S and ports lacking MOSS-specific parameters were downgraded. The "MOSS Base" version of Faceless Gods has the same file hash as another version on the same page and was excluded as a duplicate upload.
Strongly Recommended (1)
[S] BAStyle - Blue Archive Clean Cel-Shaded Style (MOSS-Image-Base 5B) · v2.0
LoRA · Version Update · Author: nekoshrimp · Published 11-13 00:21 · Downloads: 40
BAStyle v2.0: A refined version of the Blue Archive-style clean cel-shaded LoRA, focusing on facial features and image quality. It improves facial proportions and hair texture for original single characters, resulting in sharper lines and stronger contrast.
Upgraded from automatic B to S: Re-annotated training images fixed facial feature issues. Parameters, prompt templates, and workflows are complete, allowing for direct implementation.
Images

Reasons for Recommendation
- Version Update: The author states that bangs, eyes, blush, and expressions in 58 original training artworks were meticulously re-annotated. This corrects occasional outdated anime-style facial structures and oily hair textures in v1.0 single-character generations (especially with Chinese or short prompts). v2.0 features sharper line art, stronger contrast, and saturation. If you prefer the soft, low-contrast look of v1.0, you can continue using v1.0 or lower v2.0 to 0.80–0.85.
- Suitable for close-up character portraits in a lobby style, multi-character story CGs, academy buildings, and sky backgrounds. Halos are not tied to trigger words; describe their color and shape as needed. Qwen3.5-9B is frozen, training only the DiT (Rank 32 / Alpha 16). The author claims this preserves the base model's understanding of Chinese and English text and its ability to render text in images.
- Parameters are complete and include a ComfyUI workflow zip file. The full model card provides prompt templates for single-character, multi-character, and background-only scenarios, along with full validation prompts. 9 visible example images are provided.
How to Use
- Start prompts with
bastyle,followed by a complete natural language long sentence. Avoid tags likemasterpieceor1girl. For empty scenes with no people, start with2D anime background illustration. For a sky with concentric rings, addwith faint concentric rings. - Use strength 0.90–0.95 when used alone; use 0.80–0.90 when stacking with character LoRAs. Use
euler+simple, 28–32 steps, CFG 3.0. - Refer to the included workflow for settings on Flow Shift and Skip-Layer nodes.
Risks and Limitations
- The license does not allow any commercial use. Mimics commercial mobile game art styles.
- The recommended strength when stacking character LoRAs is inconsistent in the page text. Follow the included workflow.
Trigger Words: bastyle
Recommended (8)
[A] DecoSplosion · v1.0 Moss
LoRA · Port · Author: Daalis · Published 11-13 04:54 · Downloads: 34
DecoSplosion style LoRA: Builds the subject using dense decorative patterns and small elements resembling collages, paper cutouts, and stickers. The image looks like a "decoration explosion" composed of many miniature works.
Downgraded from automatic S to A: The style is novel and usage is clear, but there are no instructions specific to the MOSS version.
Images

Reasons for Recommendation
- Port: Originally a Krea 2 version. The MOSS version update note is only one sentence (a sticker battle on the site prompted the author to release a MOSS version), with no explanation of differences between the MOSS and original versions.
- Suitable for fairy tale characters, fashion portraits, decorative animals, surreal creatures, and graphic posters. The author states that subjects with interesting outlines and simple compositions best allow space for decoration.
- Strength tiers are clear: 0.5–0.8 preserves the base model's original look while adding collage patterns; around 1.0 offers the most stable style; 1.2–1.5 is more cartoonish and abstract. 8 visible example images are provided.
How to Use
- Include the trigger word
DecoDaalin the prompt. Start with strength 1.0. For a restrained look, lower to 0.5–0.8; for a weirder look, push to 1.2–1.5. - Prompts should be simple. Steps and CFG are not provided by the author.
Risks and Limitations
- The author states that results are unstable with complex prompts and may suddenly shift to a 3D feel.
- Strength tiers follow instructions from the Krea 2 era, with no MOSS-specific comparison. 3 examples are hidden.
Trigger Words: DecoDaal
[A] Deadpan Chibi Doodle · v1.0
LoRA · New Model · Author: Konan · Published 11-13 03:26 · Downloads: 54
Deadpan Chibi Doodle style LoRA: Minimalist chibi doodles with rough pencil lines, flat colors, and a pure white background, resembling an expression sticker.
Images

Reasons for Recommendation
- New model; suitable for memes, stickers, avatars, and casual chibi character thumbnails. The white background and flat colors make it easy to cut out.
- The author provides a starting sentence and a style sentence for prompts, along with a strength range of 0.8-1.3. 10 visible example images are provided.
How to Use
- Start prompts with
Minimalist cute chibi 2D doodle offollowed by the subject, then addThick textured pencil line art, flat colors, clean plain white background, expressive cartoon sticker style.. Use strength 0.8-1.3. - Steps, CFG, and sampler are not provided by the author.
Risks and Limitations
- The description is only a few lines, stating only that it was trained on MOSS. 10 examples are hidden, and the page contains mixed content.
Trigger Words: Minimalist cute chibi 2D doodle of, Thick textured pencil line art, flat colors, clean plain white background, expressive cartoon sticker style.
[A] V67 Watercolor soft · V1 MOSS
LoRA · Port · Author: angelomaiota · Published 11-13 01:55 · Downloads: 36
MOSS version of V67 Watercolor Soft: Mimics traditional watercolor—wet-on-wet blooms, cauliflower edges, cold-pressed paper texture, pigment granulation, and white highlights. The overall look is soft and low-contrast. This special edition often includes a floating glass cube or rose petal as a visual anchor.
Downgraded from automatic S to A: Usage is detailed, but the MOSS version has only 3 example images.
Images

Reasons for Recommendation
- Port: The same model card covers Z Image Turbo, Krea 2, and MOSS. The author states that the new 12 phrase trigger line activates more reliably on Turbo-style models.
- Suitable for watercolorizing landscapes, portraits, still lifes, interiors, and fantasy scenes. The model card includes multiple full-sentence example prompts.
- Starting parameters are provided: strength 0.7, guidance 5.0, using 3–4 trigger phrases. For more visible paper texture, increase strength to 0.9 and lower guidance to around 4.0. Free, with a license allowing commercial use and derivatives.
How to Use
- Prompt structure: Scene description +
(V67 Watercolor Soft:1.3)+ select 3–5 phrases from the trigger line (e.g.,soft watercolor,cold-pressed paper grain,wet-on-wet blooms,diffused daylight). The author suggests keeping prompts between 40–80 words. - Start with strength 0.7 and guidance 5.0.
Risks and Limitations
- Only 3 visible examples. The strength, guidance, and step ranges specifically listed for MOSS are only found in the table within the full model card.
- The registered trigger word is still an old full prompt (including a fishing boat scene), which is inconsistent with the new trigger line recommended in the model card. The floating cube is a fixed element in this version; if not needed, replace or remove it in the prompt.
Trigger Words: traditional watercolor painting on cold-pressed cotton paper, visible paper grain and deckled texture, wet-on-wet blooms and cauliflower edges, soft pigment bleeding into damp paper, transparent glazed layers with preserved white paper for highlights, granulation where pigment settles in paper valleys, subtle backruns and drying marks, loose expressive brushwork, muted earthy palette with a few saturated accents, soft diffused daylight, atmospheric perspective, slight color pooling at wash edges, no harsh outlines, imperfect organic shapes, hand-painted feel, a small fishing boat resting on calm water, reflections broken by water ripples, faint salt-texture speckles, (authentic watercolor medium:1.2), (not digital, not airbrushed:1.1)
[A] MOSS 12 Step Turbo (base & sft) · v9 (SFT)
LoRA · Version Update · Author: anyMODE · Published 11-13 05:21 · Downloads: 69
MOSS 12-step Turbo v9 (SFT): A rank 64 distillation acceleration LoRA. The author claims it produces realistic images in 12 steps at CFG 1.0 without quality loss.
Images

Reasons for Recommendation
- Version Update: This version is labeled SFT, while v8 (BASE) is also released on the same page. The update note only states that 8-12 steps are usable and CFG can be increased to 2.0, without explaining changes compared to v8.
- The author claims it was distilled directly from the 5.2B Base model using online teacher-student trajectory matching, compressing prompt adherence and human anatomy from 24 steps at CFG 3.0 down to 12 steps.
- ComfyUI usage is fully provided. 7 visible example images, none hidden. The page has 974 downloads and 6 comments.
How to Use
- Load the LoRA in ComfyUI with strength 1.0. Use the
simplescheduler, 8-12 steps, and CFG 1.0–2.0. - The version name is labeled SFT, intended for use with MOSS-Image SFT. If using Base, choose v8 (BASE) on the same page.
Risks and Limitations
- The author did not write an update note (only repeating recommended steps and CFG). The difference between v9 and v8 is unclear.
- "No quality loss" is the author's claim, with no comparison images or quantitative data. The sampler is not provided by the author.
[A] Dizzle's Classic Art Style (A Dabble Blend) · v1
LoRA · New Model · Author: Maridizzle · Published 11-13 07:51 · Downloads: 18
Dizzle's Classic Art Style: A warm, American mid-century magazine cover illustration style, featuring small-town daily life, soft lighting, and narrative scenes with a sense of story.
Images

Reasons for Recommendation
- New model; serves nostalgic illustrations, holiday/family scenes, and narrative images with character interactions. The author states it is best suited for everyday moments like a glance or a gesture, and warm-lit interiors.
- Full parameters: Trigger word
dz_classicfollowed bya classic painting, strength 0.9 to 1.0, approximately 30 to 50 steps, CFG 3, Euler flowmatch; the author advises using full sentences rather than stacking tags.
How to Use
- Start the prompt with
dz_classic, a classic painting, then describe the scene in full sentences; strength 0.9–1.0, 30–50 steps, CFG 3, Euler flowmatch
Risks and Limitations
- Only 4 visible examples (2 hidden), the page contains mixed content
- The origin of "A Dabble Blend" in the name is unexplained; it was just released, with almost no download feedback
Trigger Word: dz_classic
[A] Retro Low Poly - Style MOSS · v1.0
LoRA · New Model · Author Konan · Released 11-12 22:08 · Downloads 102
Retro Low Poly style LoRA: Renders images in the low-poly 3D style of the PS1 era—blocky polygons, sharp angled edges, low-resolution pixel textures, resembling game screenshots on a 240p CRT TV.
Images

Reasons to Recommend
- New model; serves needs for "intentionally aged 3D" such as retro game screenshots, PSX nostalgia posters, and low-poly character/scene concept art
- The author provides a starting phrase
A retro low poly 3d render ofalong with a full set of auxiliary descriptive words, and a strength range of 0.8–1.3 - Free download, license allows commercial use and derivatives
How to Use
- Start the prompt with
A retro low poly 3d render of, optionally appendingLow poly 3D model, chunky polygonal geometry, sharp angled edges, low-resolution pixelated texture mapping. 240p CRT video game capture. - Strength 0.8–1.3; steps, CFG, and sampler are not specified by the author
Risks and Limitations
- Only 1 visible example images (7 hidden), the page contains mixed content, making it impossible to judge topic coverage
- The description only lists a few parameter lines, with no mention of the training set or weaknesses
Trigger Words: A retro low poly 3d render of, Low poly 3D model, chunky polygonal geometry, sharp angled edges, low-resolution pixelated texture mapping. 240p CRT video game capture.
[A] MOSS_Light_and_Shadow_slider · v1.0
LoRA · New Model · Author bwy392138239 · Released 11-13 01:29 · Downloads 78
Light and Shadow Slider LoRA: Adjusts the image's light structure on a single axis—positive values darken exposure, deepen shadows, and cool/desaturate; negative values brighten shadows, warm up tones, and extend highlights.
Suitable for creating a triptych comparison of the same image at 0 / 1.0 / 2.0 to observe changes in lighting ratios.
Images

Reasons to Recommend
- New model; serves "lighting ratio fine-tuning" for dark atmospheric images: candlelight, night scenes, ruins, icons, and dark backgrounds with a single light source. You can push the same image darker or brighter without changing the prompt
- Parameters and boundaries are clearly written: recommended range -2.0 to +3.0, commonly used 1.0–2.0; the author states that exceeding +3.0 causes shadow details to collapse, with ribbons, clothing folds, and background information disappearing first
- The author's examples use the same seed and prompt, paired with
MOSS_8-steps-dmd2-lora @ 1.0, with the slider set to 0 / 1.0 / 2.0 / 3.0 for comparison; the license allows commercial use, free download
How to Use
- Load and adjust strength directly, no trigger word needed; commonly used 1.0–2.0, range -2.0 to +3.0
- When paired with a 8-step acceleration LoRA and CFG 1.0, the slider slightly alters composition: fix the seed, or use img2img denoise 0.5–0.6 to lock the framing
Risks and Limitations
- The author explicitly states it is not suitable for daylight, white backgrounds, or product shots
- Only 1 example images, stability across topics is unknown; just released, with very little download feedback
[A] Realistic Photo by Studio Atelier · MOSS
LoRA · New Model · Author 858654128784 · Released 11-12 21:02 · Downloads 158
Studio Atelier realistic studio photography LoRA: Seamless background paper in gray, taupe, or blue-gray + soft directional light, producing professional studio-style portraits, full-body fashion shots, and single-item still lifes, preserving realistic skin and fabric textures.
Images

Reasons to Recommend
- New model; serves realistic photography needs for "clean background + studio lighting" such as studio portraits (head-and-shoulders, half-body, sitting on a crate), full-body fashion shots, and single-item still lifes
- The author provides a complete prompt structure: start with
studioatelier, followed bystudio portrait/fashion studio/still life studio photograph of, then describe the subject, pose, background color, and lighting - Free download, license allows commercial use and derivatives
How to Use
studioatelier, studio portrait photograph of(orfashion studio photograph of/still life studio photograph of) + subject and pose +<color> seamless background, soft directional studio lighting, realistic skin texture- The trigger word is not registered in the page fields; write it at the beginning as instructed; strength, steps, and CFG are not specified by the author
Risks and Limitations
- 3 visible examples, 5 hidden, the page contains mixed content
- Background colors are limited to several neutral gray tones; the author does not mention performance with colored backgrounds or outdoor scenes
Worth Watching (16)
[B] Tequila (Arknights) · v1.0
LoRA · New Model · Author DSLM · Released 11-13 02:38 · Downloads 14
Character LoRA for Tequila from Arknights: Blonde hair, blue eyes, dog ears and tail, including clothing prompts for Elite 0, Elite 2 default skins, and two alternate skins.
Images

[B] Anime illustration · v2-Moss
LoRA · Ported · Author Grimm0 · Released 11-13 04:38 · Downloads 29
MOSS version of Anime illustration: A general anime illustration style between clear digital line art and soft watercolor textures.
Images

[B] Omar Dogan - Style 1.5/Pony/Illustrious/Noob/ZIT/Anima/Krea2/MOSS LORA · MOSS
LoRA · Ported · Author Konan · Released 11-13 00:17 · Downloads 88
MOSS version imitating artist Omar Dogan's style: Anime portrait style with traditional Copic marker coloring, line art, and large blocks of light and shadow.
Images

[B] V67 Film Noir · V1 MOSS
LoRA · Ported · Author angelomaiota · Released 11-13 04:18 · Downloads 50
MOSS version of V67 Film Noir: 1940s film noir style—high-contrast black and white, hard key light, venetian blind shadows, smoke, and rainy night streets.
Images

[B] Storybook Spark : Umbrium Painterly Cartoon Style · Moss v1.0
LoRA · Ported · Author umbrium2835 · Released 11-13 05:00 · Downloads 39
MOSS version of the Storybook Spark style LoRA: A three-dimensional cartoon character art style with exaggerated shapes, soft thick-painted shadows, glossy highlights, and partial outlines, sitting between cartoon characters and fantasy illustrations.
Images

[B] MOSS-Image-Base (5B) - GeminiCat / Gemini-chan Character LoRA (v1.0) · v1.0
LoRA · New Model · Author nekoshrimp · Released 11-13 01:47 · Downloads 24
Gemini catgirl (Gemini-chan, a character from the Bilibili AI girl anthropomorphism project) character LoRA: Locks in heterochromia (different colored eyes), a star-shaped hair clip and earring on the left side, an ahoge (cowlick), a bun hairstyle, and a gradient cat tail; supports outfit changes and stacking with style LoRAs.
Images

[B] Dizzle's Brushy Oil - MOSS · v1
LoRA · New Model · Author Maridizzle · Released 11-13 00:37 · Downloads 33
Dizzle's Brushy Oil: Thick oil painting style with heavy impasto, distinct brushstrokes, and palette knife textures.
Images

[B] Moss/Anima-Texture-Cool White Gradient Sheer Stockings/Cool White Flesh Thigh Texture · Moss-base-V1.0
LoRA · Ported · Author hardworkingsaltedfish · Released 11-13 00:03 · Downloads 90
Cool white stocking texture LoRA (ported from the Anima version): Primarily affects the texture of stockings and leg skin—cool white, semi-transparent gradient stockings paired with cool-toned skin, also shifting the overall art style towards a pale, cool white.
Images

[B] 2000's Cartoon Style · v1
LoRA · New Model · Author fyrtyv · Released 11-13 05:01 · Downloads 56
2000s American TV animation style LoRA: Thick outlines, exaggerated expressions, clean line art, flat colors, with separate trigger words for each individual show.
Images

[B] Comic Book Realism - CE · V01a - MossImage Base/SFT
LoRA · Ported · Author CreativeEdge · Released 11-13 05:00 · Downloads 51
MOSS version of the Comic Book Realism style LoRA: Renders images with a realistic American comic book texture, ported from multiple base model versions including Flux.1 D, Qwen, Z-Image, and Krea 2.
Images

【B】MOSS 12 Step Turbo (base & sft) · v8 (BASE)
LoRA · Version Update · Author anyMODE · Published 11-13 05:45 · Downloads 4 · Paid
BASE version of MOSS 12 Step Turbo: A distilled acceleration LoRA that, according to the author, compresses the effect of the Base model 24 steps CFG 3.0 down to 12 steps, CFG 1.0.
Image

【B】ChromaPop Anime · Moss
LoRA · Port · Author Grimm0 · Published 11-13 08:59 · Downloads 4
MOSS version of the ChromaPop Anime style LoRA: Japanese anime + thick comic book ink lines + high-contrast pop colors, producing vibrant illustrations with a screen-printed poster feel.
Image

【B】The Mechanic (Splatoon Raiders) · Moss-V1.0
LoRA · Port · Author Winnougan · Published 11-12 21:15 · Downloads 24
Character LoRA for The Mechanic from the new Splatoon game, focused on generating game characters with a 3D rendered look.
Image

【B】MOSS-Image-BlueArchive · moss-ba-r1
LoRA · New Model · Author Te2dou_Nin9 · Published 11-12 22:37 · Downloads 69
Style LoRA for the art style of Blue Archive, used to render anime character images in the game's official art style.
Image

【B】Faceless Gods · MOSS
LoRA · Port · Author ArsMachina · Published 11-12 23:28 · Downloads 2 · Paid
MOSS version of the Faceless Gods concept + style LoRA: Depicts "faceless deities"—terrifying yet sacred otherworldly gods, featuring a dark fantasy + baroque + surreal art style.
Image

【B】Krea 2 Raw|True First-Person POV|Realistic Human Eye Height First-Person View · MOSSv1.0
LoRA · Port · Author oreolove · Published 11-13 01:58 · Downloads 2 · Paid
MOSS version of the realistic human eye height first-person perspective (POV) concept LoRA: Corrects issues like "following shots from behind, third-person selfies, and unnatural camera heights," making the image look as if seen with one's own eyes.
Image

Included Only (18)
- Rui-San - MOSS · v1.0 — LoRA: Character LoRA for the original male character Rui-San: Set as a Japanese-Korean mixed-race swimmer and model, with freckles and a lean swimmer's physique.

- RLY Thot Shot - Bea (MOSS + KREA 2) · MOSS-v1.0 — LoRA: MOSS version of the fictional realistic female character LoRA "Bea" (set as a Las Vegas DJ) from the RLY Thot Shots series.

- alterkyon style lora | Moss · v2.0 — LoRA: v2.0 of the alterkyon style LoRA: An anime style LoRA named after alterkyon, where the author states the style was not fully trained.

- Cyberboi - Art style · MOSS IMAGE v2.0 — LoRA: The second MOSS version of the Cyberboi style LoRA, originally from the Anima series; the page does not describe the art style itself.

- MOSS-Image — Fast Text-to-Image Workflow (ComfyUI) · v1.0 — Workflow: MOSS-Image text-to-image ComfyUI workflow: The author claims it is built according to the official template, with the base model, text encoder, VAE, and default parameters configured, ready to use upon import.

- One-piece Competition Swimsuit for Moss-Image · v1.0 — LoRA: One-piece competition swimsuit clothing LoRA: Dresses characters in one-piece competition/diving material swimsuits, can be stacked with character or style LoRAs.

- Alisa | MOSS / Krea 2 / · v2.0 MOSS — LoRA: MOSS version of the character LoRA for the original realistic female character Alisa, where the author states this version features a new look: round-framed glasses, side braid, and subtle facial feature adjustments.

- Bombchu Bowling Lady 2026 (ボムチュウボウリングのお姉さん2026版) · v1.0 — LoRA: Character LoRA for the Bombchu Bowling Lady from The Legend of Zelda: Ocarina of Time (2026 remake style).

- xSerina · MOSS Image — LoRA: xSerina: Character LoRA for a fictional young blonde woman, where the author states all training images are AI-generated and do not correspond to a real person.

- xTashia · MOSS Image — LoRA: xTashia: Character LoRA for a fictional African-American woman (labeled as style, but actually a character), where the author states all training images are AI-generated.

- Takeda_Hiromitsu style · moss_v1.0 — LoRA: Style LoRA imitating the art style of manga artist Takeda Hiromitsu.

- Ratchet (Tools of Destruction) (PS3) · MOSS Image — LoRA: Character LoRA for Ratchet, the Lombax protagonist from the PS3 game Ratchet & Clank: Tools of Destruction, focused on a 3D game rendering style.

- Telepurte · Moss — LoRA: Style LoRA imitating the art style of webcomic artist tepurtle, focusing on their character nila and the faceless male figures in their work, with switchable black and white.

- Nicole (Model IA) [MOSS / Krea2 / SD] · MOSS v1.0 — LoRA: Character LoRA for the fictional woman Nicole (oval face, grey-green eyes, dark brown center-parted long hair), where the author states it does not correspond to a real person.

- Lily Wuthering Waves MOSS Image · v1.0 — LoRA: Character LoRA for Lily from Wuthering Waves, with a trigger string consisting of a long list of appearance tags.

- KRAST-Moss-0.2 (base) · v0.2 — UNet: KRAST-Moss 0.2: A realistic + aesthetic-oriented base model (full UNet weights) created by merging 7 self-trained LoRAs by the author onto MOSS Base.

- IllustraGlam for Moss · v1.0 glamour — LoRA: IllustraGlam: Described by the author as a "stylized, glamorous" illustration style LoRA.

- Kigurumi Onesie: Cat [Krea2/MOSS] · MOSS_SFT_v1.0 — LoRA: MOSS SFT version of the Cat Kigurumi Onesie clothing LoRA, dressing characters in a cat-ear hooded onesie.

Data: Civitai Public API (Base Model MOSS Image), inclusion window 11-12 20:31 → 11-13 09:01 (Beijing Time). Recommendations reviewed by editors; download counts only reflect attention.