MOSS Image Morning Report 2026-11-02 | Top Pick: leorlontyan style (Anima/Illustrious/MOSS)

Zhenzhen's AI Radar · MOSS Image · 2026-11-02 Morning Report

This issue features 69 new versions: Strongly Recommended 1 · Recommended 7 · Worth Watching 28 · Listed Only 33

Editor's Note This issue reviews 74 new MOSS versions. The top recommendation is the MOSS retrained version of artist leorlontyan's style: warm golden backlighting, soft outlines around character edges, and anime group illustrations with exaggerated expressions. Prompt templates, strength settings, and a complete set of sampling parameters are all provided, maintaining an S rating. Other highlights: a makeup slider that can be positive or negative, a black-and-white line art LoRA requiring no trigger words (upgraded from B to A), 2.5D isometric game scenes, a university party scene with detailed prompt writing, retro skeletal anatomy scientific line art, and MOEMOSS-Base, a full model aimed at vibrant anime colors. Several style or ported versions in the automatic S/A list lacked MOSS-specific parameters or style descriptions and were downgraded to B/C; standard character and person LoRAs (including 7 different characters from the Mr Piglet series) were rated C; two re-uploads with renamed files identical to the old model SHA256 were excluded.

Strongly Recommended (1)

[S] leorlontyan style (Anima/Illustrious/MOSS) · MOSS - v1.0

LoRA · Port · Author Mike_832 · Published 11-02 03:30 · Downloads 32

The MOSS version of the leorlontyan style LoRA: warm golden backlighting, a soft yellow-green outline around character edges, olive and brown tones, sparkle effects, and characters with large eyes, wide smiles, and blushing cheeks.

Maintaining S: The MOSS version includes separate training instructions and complete sampling parameters, the prompt template is ready to copy, and the license is permissive; the only concern is the imitation of a real artist's style.

Images

Reasons for Recommendation

• Port: Previous versions existed for Anima v2 and Illustrious, all three sharing the trigger word @lrlt; the MOSS version was separately retrained on MOSS-Image Base, switching to natural language prompts instead of Danbooru tags.
• Serves users who want warm-toned anime group illustrations with exaggerated expressions in MOSS; the author states the style primarily affects characters, while backgrounds retain more of MOSS's native appearance, suitable for multi-person daily scenes like dining, playing music, or cooking.
• The MOSS version parameters are very comprehensive: prompt templates, negative prompts, strength, sampler, scheduler, steps, CFG, shift, skip layers, and resolution are all provided, noting the use of SFT for image generation; 9 visible examples; free with a fully open license; this version has 32 downloads, with cumulative downloads for the entire model at 340 (indicating popularity only).

How to Use

• Load leorlontyan_moss_v1.safetensors (71 MB) with strength 1.0; use MOSS-Image SFT for generation (the author tested on SFT int8), not Base.
• Start prompts with "@lrlt.", then write full sentences describing "who, doing what, where, lighting, and mood"; tags and style words are not needed; negative prompt: sensitive content, worst quality, bad anatomy, watermark, jpeg artifact, signature.
• Sampling: er_sde + simple, 25 steps, CFG 3, shift 5, enable skip layer (author specifies layer 22), 1024×1536, do not use high-res fix.
• Explicitly requesting smiles, laughter, or blushing in the prompt aligns better with the style; do not mention signatures in the prompt, otherwise MOSS will draw them.

Risks and Limitations

• A style LoRA named after a real artist; the author declares it is not official, training images are sourced from this artist's works, and commercial use should consider controversies regarding style imitation.
• The parameter excerpt in data.json (DPM++ 2M SDE, CFG 5, 832×1248) is for the Anima version; MOSS version parameters are in the MOSS-Image section of the full model card; the author states the MOSS version release corresponds to step 1400 out of 2400 steps, trained on 101 images, and is not a final version.
• The author notes that hands tend to tangle in fighting or crossed-weapon scenes, requiring re-generation.
• 1 sensitive-level examples are hidden; this entry was not run.

Trigger Words: @lrlt

Recommended (7)

[A] Vintage Skeleton Anatomy - Scientific Line Art · v1

LoRA · New Model · Author app1ecore · Published 11-02 06:43 · Downloads 25

A vintage anatomical diagram style LoRA: generates black-and-white skeletal scientific line art without shading or coloring, resembling copperplate engravings from old biology textbooks; capable of drawing both human and animal skeletons.

Downgraded from S to A: The subject is specific and prompts are easy to write, but sampling parameters do not appear MOSS-specific, step counts are missing, and training data comes from commercial atlases.

Images

Reasons for Recommendation

• New Model: Turns the very specific technique of "old-style scientific plates" into a LoRA; the author provides three example prompts: basic, specified animal (frog dorsal view), and human standing pose.
• Provides strength and CFG ranges; all 5 visible examples include prompts; free with a license allowing image output and rental; approximately 4 hours since release with 25 downloads (indicating popularity only).

How to Use

• Load vintageSkeletonAnatomy_mossv1.safetensors (138 MB) with strength 0.6–0.8, the author suggests starting from 0.7.
• Start prompts with "line art, scientific illustration, anatomy, vintage engraving, skeleton", then describe the subject and perspective, such as "skeleton of a frog, dorsal view, black and white", or add "isolated on white background".
• CFG 5–7, samplers listed are DPM++ 2M Karras or Euler a; step count not provided by the author.

Risks and Limitations

• The recommended samplers and CFG ranges resemble generic SD templates; the author did not state if they were tuned for MOSS; step count is missing.
• Training images are taken from the "Skeletons" atlas published by Pepin Press; the author claims no intent to infringe, but the source is a commercial publication.
• 14 sensitive-level examples are hidden, far more than the 5 visible examples; this entry was not run.

Trigger Words: line art, scientific illustration, anatomy, vintage engraving, skeleton

[A] Moss/Anima-style-Dreamy Soft Light (nai5 fifth) · Moss-base-v1.0

LoRA · Port · Author hardworkingsaltedfish · Published 11-02 01:06 · Downloads 72

The MOSS version of the "Dreamy Soft Light" anime style LoRA: healing dreamy lighting, bright yet soft pastel color schemes, and a Japanese light novel illustration texture.

Maintaining A: Style descriptions, trigger words, strength, and steps are specifically written for the MOSS version, but example prompts are few and training data sources are vague.

Images

Reasons for Recommendation

• Port: A previous Anima version existed; the MOSS version was trained on MOSS-Base using the same training set, with unchanged trigger words; the author states that character texture and lighting in the MOSS version are significantly better than in the Anima version.
• Serves users creating Japanese light novel cover or illustration-style character images; the author provides a set of auxiliary prompts—use all when standalone, only use the trigger word when stacking multiple LoRAs.
• Strength and steps have MOSS-specific recommendations; free; this version has 72 downloads, with cumulative downloads for the entire model at 407 (indicating popularity only).

How to Use

• Load Moss-Style-nai(fifth).safetensors (138 MB), write "@softglow_anime" in the prompt; when used standalone, add "dreamy healing atmosphere, vibrant yet soft pastel colors, polished high-quality japanese light novel illustration style".
• Strength 1 for standalone use; the author suggests 0.5-0.8 or 0.8-1.3 when stacking with other LoRAs; 25-35 steps for daily use.
• The author advises against using acceleration LoRAs if possible, as the style may be partially lost; CFG and sampler not provided by the author.

Risks and Limitations

• Only 1 of the 6 examples include prompts.
• The Anima version description states training materials are from the public internet; "nai5" in the model name may indicate training images come from NovelAI generation, but the author did not clarify.
• License allows only on-site rental, prohibits merging, and requires attribution; 4 sensitive-level examples are hidden; this entry was not run.

Trigger Words: @softglow_anime

[A] MOEMOSS-Base · Base

Checkpoint / Base Model · New Model · Author Minokai · Published 11-01 23:10 · Downloads 40

An anime-oriented full fine-tuning checkpoint for MOSS: the author states it was trained on 10,800 high-rated anime images, focusing on character standing poses, portraits, and illustration scenes with vibrant colors.

Downgraded from S to A: Base model fine-tuning, complete parameters, and permissive license, but few visible examples and unclear data sources.

Images

Reasons for Recommendation

• New Model: A complete base model rather than a LoRA, with 800 images added by the author based on personal preference; the author claims this mix works well with various style LoRAs and plans to continue expanding data based on it.
• Serves users who want MOSS to default to colorful anime illustrations before stacking style LoRAs; examples for the same author's Creamy Peach style LoRA were generated on top of this model.
• Provides complete sampling parameters, example images use only this model without LoRAs; available in INT8 and BF16 files; free with a fully open license; approximately 12 hours since release with 40 downloads (indicating popularity only).

How to Use

• Download moemossBase_base_bf16 (9.5 GB) or the roughly half-sized moemossBase_base_int8, and replace the MOSS base model for loading.
• Author's example parameters: Euler + sgm_uniform, 50 steps, CFG 3; the author states it can be paired with Turbo LoRA for acceleration (specific file not provided).

Risks and Limitations

• Only 4 visible examples, 2 sensitive-level examples are hidden.
• Training image sources are not specified, only described as high-rated anime images.
• First version; the author states data will be expanded later; this entry was not run.

[A] wakfo-iso |isometric|game asset map · v1.0

LoRA · New Model · Author p0816244617 · Published 11-01 22:43 · Downloads 53

A 2.5D isometric perspective game scene / game asset LoRA: trained on 105 isometric game scene images, capable of directly drawing isometric map scenes and individual assets.

Maintaining A: Specific use case, all examples public, permissive license, but no sampling parameters provided and training image sources are questionable.

Images

Reasons for Recommendation

• New Model: The author's goal is to batch-create game assets using open-source models, choosing MOSS for its ability to precisely control composition via JSON structured prompts.
• Serves indie game prototypes, isometric maps, level concepts, and asset sketches; 10 visible examples, no hidden examples.
• Free with a fully open license (allows image output, sale, merging); approximately 12 hours since release with 53 downloads (indicating popularity only).

How to Use

• Load wakfu-iso-scene_c1-st1000.safetensors (74 MB), use the prompt "wakfu-iso", and combine with MOSS's JSON prompt to arrange tile and object positions
• Strength, steps, CFG, and sampler are not provided by the author

Risks and Limitations

• The trigger word is not registered in the version information, only mentioned in the description (the model name is misspelled as wakfo)
• The name refers to the game "Wakfu"; the source of the training images is not specified and may come from the game's art assets, so be mindful of copyright for commercial assets
• The author describes it as experimental, with no sampling parameters; this entry has not been tested

【A】College Party — Red Cups, Beer Pong, Loser Strips | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 21:05 · Downloads 125

A concept LoRA for American college house party scenes: string-lit living rooms, red plastic cups, pizza boxes and snacks, kitchen islands crowded with liquor bottles, backyard barbecues, and city night skyline rooftops, with crowds packed onto sofas, around tables, and at doorways.

Maintain A: The scene elements and prompt writing style are clearly explained and easy to follow, but there are no sampling parameters, and hidden examples make up the majority.

Images

Reasons for Recommendation

• New model: A new entry in the same author's scene concept series, turning a specific scene into a template. The author wrote the prompt order and three camera angles (end of the beer pong table, within a circle of friends sitting on the floor, over the shoulders of the crowd)
• Suitable for realistic narrative illustrations or snapshot-style images for party and campus themes; the author suggests placing a specific object in the foreground (a row of red cups, a jack-o'-lantern, liquor bottles) to make the image look more like a real party snapshot
• All 4 visible examples include prompts; free, with a fully open license; already has 125 downloads, the most in this group (only indicates popularity)

How to Use

• Load collegepartyconcept.safetensors (142 MB), start the prompt with "collegeparty," (the author emphasizes that the trigger word is collegeparty, not the file name)
• First write the camera angle, then the lights, red cups, sofa, and kitchen, and finally each person's appearance, age (in the format "an adult man aged 23"), position, action, and expression; the author requires all characters to be written as adults
• Strength, steps, CFG, and sampler are not provided by the author

Risks and Limitations

• The trigger word is not registered in the version information, only mentioned in the description
• No sampling parameters; this entry has not been tested

【A】Slider: Makeup · MOSS-base v1.0

LoRA · New Model · Author NerissG · Published 11-02 01:39 · Downloads 82

A makeup intensity slider LoRA: positive weights make makeup lighter, closer to bare-faced, while negative weights add heavy makeup; the author states that both male and female, illustration and photo data were involved in training.

Maintain A: The function is single and clear, with the direction and limitations well explained, allowing for positive/negative comparisons with the same prompt; however, there are only 1 visible examples and no sampling parameters, which is not enough for an S rating.

Images

Reasons for Recommendation

• New model: Turns "how heavy the makeup is" into a continuous knob that can be positive or negative, without changing the prompt; the placeholder words in the training captions are "natural looking, barefaced" for positive and "full glam, heavy makeup" for negative
• Serves the need to uniformly adjust makeup intensity in portrait generation, such as creating bare-faced / light makeup / heavy makeup comparisons for the same character, usable for both realistic and illustration styles; the file is only 18 MB
• The author proactively wrote the limitations: adding makeup tends to make colors more vivid, while removing makeup tends to make them cooler; when the negative value exceeds -1, the character's appearance becomes exaggerated and sexualized. Free, with a fully open license but requires attribution; published about 9 hours ago with 82 downloads (only indicates popularity)

How to Use

• Load MakeupSlider_Moss-Base_v1.safetensors (18 MB), no trigger word; positive values reduce makeup (natural / bare-faced), negative values add makeup (heavy makeup)
• It is recommended not to exceed -1 in the makeup-adding direction; specific strength ranges, steps, CFG, and sampler are not provided by the author; the file name indicates it was trained based on MOSS Base

Risks and Limitations

• Only 1 visible examples, 6 sensitive-level examples are hidden
• Color cast: adding makeup tends to be more vivid, removing makeup tends to be cooler
• The author states that this version reversed the positive and negative directions, so pay attention to signs when combining with other sliders; this entry has not been tested

【A】MOSS Line Art Style · v1.0

LoRA · New Model · Author manakan · Published 11-01 21:36 · Downloads 73

A LoRA that turns MOSS outputs into black and white line art: the author claims it does only one thing, removing color and leaving only lines; it works without a trigger word, but writing "line art" or "sketch" yields better results.

Upgraded from B to A: The purpose is clear in one sentence, no trigger word needed, fully open license, and visible examples exist, only lacking parameters.

Images

Reasons for Recommendation

• New model: Solves the need for "wanting clean line art, but the base model always adds color," usable for generating coloring book drafts, sketch references, or as a starting point for subsequent coloring
• Low barrier to entry: No trigger word needed, the author claims adding "line art" or "sketch" to the prompt yields better results; the file is only 63 MB (rank 16)
• Has 2 visible examples (another 2 are hidden), all with prompts; free, with the most permissive license in this group, allowing output, rental, and sale of fusions; published about 13 hours ago with 73 downloads (only indicates popularity)

How to Use

• Load moss_lineart_v1_rank16.safetensors (63 MB · BF16), no trigger word needed, writing "line art" or "sketch" in the prompt is more stable
• Strength, steps, CFG, and sampler are not provided by the author, you can first refer to the prompts attached to the example images

Risks and Limitations

• The description is only two sentences, with no sampling parameters or strength suggestions
• The author's original words are "desaturate the image into line art," without specifying whether it can be combined with img2img / editing workflows to convert existing images into line art, currently understood as a text-to-image style LoRA
• Only 2 visible examples, the line quality for complex scenes and subjects other than characters is unknown; this entry has not been tested

Worth Watching (28)

【B】Chamuring - Artist Style · MOSS-Image

LoRA · Ported · Author Azriaelis · Published 11-01 23:52 · Downloads 39

The MOSS version of the artist Chamuring's style LoRA; the author provides no textual description of the style, but based on tags, it appears to be thick-painted, sketch-like beautiful male illustrations.

Images

【B】Impasto - MOSS Image v1 · v1.0

LoRA · New Model · Author inpaint · Published 11-02 02:48 · Downloads 39

An impasto / finger painting style LoRA: the image features obvious paint stacking brushstrokes, primarily used by the author for anime characters.

Images

【B】Fantasy Graphite Sketch · v11 - MOSS

LoRA · Version Update · Author Fatbuns · Published 11-01 23:10 · Downloads 38

The second MOSS version (v11) of the "Fantasy Graphite Sketch" fantasy pencil sketch style LoRA; the author claims this version is closer to the original dataset.

Images

【B】【MOSS-image】Vesna and Danica Two-Character LoRA · Standard

LoRA · New Model · Author wiehhg_37 · Published 11-02 08:54 · Downloads 12

A MOSS two-character LoRA containing two characters (Vesna and Danica) in one LoRA, capable of generating single-person images or images with both characters together; the author claims that when prompts are detailed, the features of the two characters rarely bleed into each other.

Images

【B】NexBlend 03 Asian Realistic LoRA - MOSS Image · v1.0

LoRA · New Model · Author NexBlend · Published 11-02 05:26 · Downloads 43

An East Asian realistic portrait LoRA leaning towards Korean aesthetics: the author claims it enhances Korean facial features and can reproduce scenes such as Korean city streets, signs, hanok houses, and Joseon dynasty clothing.

Images

【B】Moss-image - Arsh illustration | Style · v2.0

LoRA · Version Update · Author GreatWhiteWolf · Published 11-02 00:21 · Downloads 35

The MOSS v2.0 of the "Arsh illustration" male-oriented illustration style LoRA: trained with 180 images, the author claims this version's style and details are closer to the original.

Images

【B】Ixy (MOSS) | Artist styles · v1.0

LoRA · New Model · Author lerosay · Published 11-02 00:09 · Downloads 80

A LoRA imitating the style of artist Ixy: flat coloring, thin and clear lines, cute moe-style characters.

Images

【B】Anon 2-okunen style | [MOSS | Anima | Illustrious] · moss_1.0

LoRA · Ported · Author NoiseWriggler · Published 11-02 05:25 · Downloads 47

The MOSS version of the manga LoRA imitating the style of doujinshi artist Anon 2-okunen: trained on MOSS-Image-Base and tested on MOSS-Image-SFT.

Images

【B】Envy Anime Oil MOSS/SDXL 01 · MOSS v1.0

LoRA · Ported · Author _Envy_ · Published 11-02 05:54 · Downloads 35

The MOSS version of the "Envy Anime Oil" anime oil painting style LoRA: the original SDXL version was used to give anime characters themselves oil painting brushstrokes, not just the background.

Images

【B】Moss-image - Betm illustration | Style · v2.0

LoRA · Version Update · Author GreatWhiteWolf · Published 11-02 01:42 · Downloads 48

The MOSS v2.0 of the "Betm illustration" male-oriented illustration style LoRA: trained only on original illustration images, without anime assets, the author claims it can enhance the realistic texture of skin.

Images

【B】Minokai Styles - Creamy Peach · v1.0

LoRA · New Model · Author Minokai · Published 11-01 23:25 · Downloads 53

Author Minokai's personal style LoRA "Creamy Peach": leaning towards 2D illustration with a touch of 2.5D, rich colors, strong chromatic aberration effects, reddish skin, focusing on character portraits with expressive faces.

Images

【B】MiaoMiao Harem MOSS — ComfyUI Workflow (8GB) · v1.0

Workflow · New Model · Author tinytensor · Published 11-02 10:14 · Downloads 4

A ComfyUI text-to-image workflow configured for the community base model MiaoMiao Harem MOSS (INT8 version). The author tuned the parameters on a laptop with 8GB of VRAM and recorded the generation times; it does not include new weights.

Images

【B】Sakalgath Style [Anima/Krea2/MOSS] · MOSS v1.0

LoRA · Port · Author Imagine__That · Published 11-01 22:07 · Downloads 68

The MOSS version of the "Sakalgath" art style LoRA: the author describes it as a low-saturation, thick-line manhwa (Korean comic) style.

Images

【B】Mrpopo's clarity-moss-image · Mrpopo Hires Moss further

LoRA · Version Update · Author Mrpopo714 · Published 11-02 10:33 · Downloads 1 · Paid

The third version of the realism enhancement / high-resolution LoRA "Mrpopo's clarity": retrained using only 2048 resolution on the same dataset. The author claims significant improvements in close-up skin and lighting, targeting outputs of approximately 4MP.

Images

【B】MOSS Shiny Anime · v1.0

LoRA · New Model · Author _Envy_ · Published 11-02 08:25 · Downloads 15

A "cute, glossy" anime style LoRA. The author only wrote this one sentence and mentioned they didn't know how to name it.

Images

【B】Moss-image - Betm Anime | Style · v1.0

LoRA · New Model · Author GreatWhiteWolf · Published 11-02 00:19 · Downloads 38

A flat anime style LoRA: the author claims it was trained on 330 pure anime images, deliberately excluding complex and detailed illustration assets; the tags lean towards male characters.

Images

【B】Onono Imoko Style (Dohna Dohna) · v1.0

LoRA · New Model · Author Minokai · Published 11-02 00:01 · Downloads 44

An anime LoRA imitating the style of onono imoko, the original artist for Dohna Dohna: the author claims its characteristics are bold color block compositions and cute female characters, with slight screen-door artifacts appearing at high strength.

Images

【B】XI410_in_MossImage · v1.0

LoRA · New Model · Author 1822510823838 · Published 11-02 01:02 · Downloads 80

An anime LoRA imitating the style of pixiv artist XI410. The author claims it has almost no effect on real-life photos.

Images

【B】Microsoft Paint 98 - amateur pixel comic style · v1.0

LoRA · New Model · Author jpie · Published 11-01 21:58 · Downloads 44

A LoRA imitating an amateur pixel comic style reminiscent of "Windows Paint": the author jokingly claims it transforms MOSS images into the style of an amateur comic artist "found on an old floppy disk," and released the training set with subtitles.

Images

【B】Lunar Etching · MOSS Image

LoRA · Port · Author Wackiii · Published 11-01 23:28 · Downloads 54

A monochrome retro etching style LoRA: the author claims it combines fine ink line engraving, celestial symbols, Art Nouveau lines, and parchment texture, commonly featuring cross-hatching, Gothic architecture, and large eclipsed moons.

Images

【B】Minokai Styles - Grapefruit · v1.0

LoRA · New Model · Author Minokai · Published 11-02 01:00 · Downloads 44

The "Grapefruit" anime style LoRA: the author claims it centers on vibrant colors and expressive characters, being flatter and having more angular faces than the same author's Creamy Peaches.

Images

【B】Super Tall Girl with Long Legs/长腿高挑女孩 · Moss V1.0

LoRA · Port · Author aki_11679 · Published 11-02 00:36 · Downloads 44

The MOSS version of a tall, long-legged female body type LoRA: the author claims it doesn't simply scale up the character, but changes body proportions (increasing head-to-body ratio, lengthening legs) to represent "tallness."

Images

【B】moemoe piano style · v1.0

LoCon · New Model · Author ibara0608 · Published 11-01 21:44 · Downloads 49

A "slippery" 1girl anime style LoCon imitating the NovelAI v5 artist tag moemoepiano. It must be used with the DMD2 acceleration LoRA and the trigger word "Separated long eyelashes."

Images

【B】moss winged eyeliner · v1

LoRA · New Model · Author hamstornado · Published 11-01 23:07 · Downloads 8 · Paid

A MOSS concept LoRA that adds "winged eyeliner" makeup to characters.

Images

【B】moss bee-stung lips · v1

LoRA · New Model · Author hamstornado · Published 11-01 22:03 · Downloads 2 · Paid

A MOSS concept LoRA that generates "bee-stung lips" (full, slightly pouty lips). The author claims it is deliberately kept natural, avoiding exaggerated hyaluronic acid filler effects.

Images

【B】Double below knee amputee · Moss Image

LoRA · Port · Author heidisrsen520 · Published 11-02 01:19 · Downloads 30

A MOSS concept LoRA for depicting female characters with "bilateral below-knee amputations," supplementing disability body features that base models typically fail to draw.

Images

【B】moss hijab · v1

LoRA · New Model · Author hamstornado · Published 11-01 22:40 · Downloads 2 · Paid

A MOSS concept LoRA that puts a hijab on characters. Prompts can be used to switch between various wrapping styles and under-cap colors.

Images

【B】[Moss] Melailai Style · v1.0

LoRA · New Model · Author aki_11679 · Published 11-02 00:22 · Downloads 0 · Paid

A MOSS LoRA replicating the style of artist Melailai. The author provided full weights, sampler, steps, CFG, and SLG parameters, and warned against using er_sde.

Images

Included Only (33)

• Enola - a Fake Person · Moss — LoRA: The MOSS version of a realistic portrait LoRA for the fictional character "Enola": the author claims it is a non-existent blonde woman created by mixing multiple faces. This version adjusts the age to late twenties to early thirties.
• MOSS-FET · v0.2 — Base Model / Checkpoint: The second version of an experimental checkpoint for full fine-tuning of MOSS: continued training on a small dataset of a different art style based on v0.1, leaning towards "semi-realistic." The author self-declares the results are poor and releases it only for testing.
• akipeko style · v1.0_moss_base — LoRA: The MOSS-Base version of the akipeko thick-paint style LoRA. The author claims it was "distilled" from another akipeko style model on Civitai.
• Moss-image - conndycute | Style · v1.0 — LoRA: The MOSS version of the "conndycute" male-oriented style LoRA. The author claims it was trained on 348 curated images; the style is not described.
• Moss-image - MilkteaSfisdis | Style · v1.0 — LoRA: The MOSS version of the "MilkteaSfisdis" male-oriented style LoRA. The author claims it was trained on 120 images; the style is not described.
• Valeria Gross [MOSS] · v1.0 MOSS — LoRA: A character LoRA for an AI-generated fictional realistic mature female character (the author claims approximately 38–42 years old), requiring no trigger word.
• Rampage - Art Style · MOSS Image v1.0 — LoRA: The MOSS version of the "Rampage" art style LoRA. The author does not describe the style, only providing the base model, auxiliary LoRA, and sampling parameters for reproduction examples.
• Misty: Tomboyish Mermaid | カスミ (Pokemon) [MOSS] · v1.0 — LoRA: A character LoRA for Misty (カスミ) from Pokémon, MOSS version.
• Flux2jm-vae · x1 — VAE: A VAE modified from the official Flux.2 Dev VAE; the author claims it includes "darkness adjustments" and is compatible with Flux.2, MOSS, and similar models.
• Dragonmaid ドラゴンメイド (Yu-Gi-Oh! 遊戯王) · moss_image_v1.0 — LoRA: MOSS version of a collection LoRA for the "Dragonmaid" card series characters from Yu-Gi-Oh!, containing trigger words for 7 characters in a single file.
• Better Compression Shirts · v1.0 — LoRA: A tight compression clothing LoRA: the author claims it can render more detailed sports compression shirts that reveal muscle definition, supporting both short-sleeve and long-sleeve styles.
• Stormy Night Moss Turbo · 1.0 — Base Model / Checkpoint: A complete MOSS Turbo base model in a semi-realistic digital illustration style: the author highlights glossy skin, clear lines, reflective leather and metal, and high-contrast dramatic lighting, with the trigger word "macasy style"; access requires payment or membership via the author's Ko-fi.
• shortstack (Ver. Anima, Illustrious, flux, NoobAi) · vMoss_stf — LoRA: MOSS version of the "shortstack" body type concept LoRA; the filename indicates it was trained on MOSS-Image-SFT.
• WhirlWind【SD1.5+2.5DStyle】 · MOSS v100beta — Base Model / Checkpoint: Beta version of the "WhirlWind" 2.5D anime style base model for MOSS: a full 5.0 GB set of INT8 weights, which the author claims was retrained from scratch for MOSS, aiming for the 2.5D anime illustration feel of the SD1.5 era.
• Body Paints Moss Image by Sarcastic TOFU · Body_Paints_Moss_v1 — LoRA: A MOSS LoRA that adds body paint (colored pigments, camouflage, wet/dry paint effects) to characters, with the trigger word Body_Paints.
• Great Fairy (Ocarina of Time) [MOSS & ANI & IL] · MOSS — LoRA: MOSS version of the Great Fairy character LoRA from The Legend of Zelda: Ocarina of Time, also including a trigger word for the in-game "Milk Bar" scene.
• Lindvern Sunbringer (Zenless Zone Zero) · v1.0 — LoRA: MOSS character LoRA for Lindvern Sunbringer from Zenless Zone Zero, including the base outfit and a set of alternate outfits, each with independent trigger words.
• Civchan - Stewi0001 Edition · MOSS v1.0 — LoRA: MOSS version of a third-party character LoRA (stewi0001 edition) for Civitai's mascot "Civchan".
• Mr Piglets Oldies and brats Moss image · Lux41 — LoRA: A female character numbered "Lux41" from Mr_Piglet's MOSS character LoRA collection, with the trigger word being the version name.
• Mr Piglets Oldies and brats Moss image · Lux40 — LoRA: A female character numbered "Lux40" from Mr_Piglet's MOSS character LoRA collection, with the trigger word being the version name.
• Mr Piglets Oldies and brats Moss image · Lux39 — LoRA: A female character numbered "Lux39" from Mr_Piglet's MOSS character LoRA collection, with the trigger word being the version name.
• Mr Piglets Oldies and brats Moss image · Lux38 — LoRA: A female character numbered "Lux38" from Mr_Piglet's MOSS character LoRA collection, with the trigger word being the version name.
• Sergal · v1.0 MOSS-Image — LoCon: A MOSS LoKr for drawing the furry species "Sergal" (anthropomorphic creatures with long pointed snouts and pointed ears).
• xTerri · MOSS Image — LoRA: A MOSS character LoRA for a blonde female virtual character, which the author claims is entirely AI-generated.
• Mr Piglets Chavettes and Essex women for Moss · Chav5 — LoRA: The fifth version, Chav5, from Mr_Piglet's "Chavettes and Essex women" MOSS character collection, featuring women dressed in British chav / Essex style.
• Mr Piglets Oldies and brats Moss image · Lux37 — LoRA: A female character numbered "Lux37" from Mr_Piglet's MOSS character LoRA collection, with the trigger word being the version name.
• Mr Piglets Chavettes and Essex women for Moss · Chav4 — LoRA: The fourth version, Chav4, from Mr_Piglet's "Chavettes and Essex women" MOSS character collection.
• onineko moss image · v1.0 — LoRA: A MOSS style LoRA named "onineko", with the description consisting only of a trigger word.
• moss appleshape body · v1 — LoRA: A MOSS body type concept LoRA that changes the character's body shape to "apple-shaped" (wider waist and abdomen).
• Lino · v1.0 — LoRA: A MOSS character LoRA named "Lino", with a completely blank model card.
• MOSS Image GGUF Q8 · v1.0 SFT — Base Model / Checkpoint: GGUF Q8 quantized version of the MOSS Image SFT base model (5.3 GB), which is merely a format conversion and does not add new capabilities.
• d1m style · v1.0 MOSS-Image — LoCon: A flat-shaded furry style LoKr ("d1m style"), with the description consisting only of "Flat shading, furry style".
• MOSS Image GGUF Q8 · v1.0 Base — Base Model / Checkpoint: GGUF Q8 quantized version of the MOSS Image Base model (5.3 GB), accompanying the SFT version on the same page, which is merely a format conversion.

---

Data: Civitai public API (base model MOSS Image), collection window 11-01 20:31 → 11-02 10:45 (Beijing Time). Recommendations are reviewed by editors; download counts only reflect attention.