Zhenzhen's AI Radar · MOSS Image · 2026-11-22 Evening Report
This issue features 44 new versions: Strongly Recommended 2 · Recommended 8 · Worth Watching 19 · Included Only 15
Editor's Note This issue reviewed 48 new MOSS versions, with 2 S, 9 A, 20 B, and 17 C ratings, and no exclusions. Top pick: the wrad art style, a NovelAI-style animation style where a single LoRA switches between chibi and standard head-to-body ratios via prompts; base model, acceleration LoRA, steps, and CFG are all provided. Another S-rated item is a group photo control LoRA that uses per-person templates to specify headcount, positioning, and each person's appearance, upgraded from A to S. Other highlights include a trigger-word-free skin modifier slider, MOSS merge workflows, Sequence7 multi-LoRA workflow v1.2, and Dizzle Galaxy. Auto-S rated Korean beauty portraits, Grimm realistic oil paintings, and Crocell backgrounds were downgraded to A due to incomplete MOSS parameters; character LoRAs in the same batch were mostly downgraded to B or C due to few examples or copyrighted characters.
Strongly Recommended (2)
[S] wrad style | NAI | NovelAI | Chibi [MOSS] · v1.0
LoRA · New Model · Author len · Published 11-22 13:35 · Downloads 43
A NovelAI-style animation LoRA named after @wrad: capable of generating both chibi (large head, small body) and standard head-to-body ratio characters, switching via prompts; trained directly for MOSS.
Run the same seed with "chibi, chibi only" and "tall female" prompts to directly verify if the head-to-body ratio switch is clean.
Images

Reasons for Recommendation
- New model: A single LoRA switches between two ratios using prompts—writing "chibi, chibi only" enhances the chibi style, while removing it and writing "tall female" returns to standard animation proportions, serving chibi avatars, stickers, memes, and standard animation character art.
- MOSS parameters are the most complete: Base model + the author-specified 8-step DMD acceleration LoRA, with weights, steps, CFG, and training resolution all documented.
- Visible 8 examples (another 2 sensitive-level examples hidden), 43 downloads and 5 likes within 7 hours of release; the author claims it also works for furry themes.
How to Use
- Trigger word @wrad, LoRA weight 1.0; base model MOSS Image Base, strongly recommended to stack lodestones/MOSS-Image-SFT-8-steps-dmd (dmd_latest, strength 0.7 or 1.0).
- 8 steps, CFG 1.0; trained at 1024 × 1024, the author claims higher resolutions are sharper but low resolutions also work; add "chibi, chibi only" for chibi style, or write "tall female" for standard proportions.
Risks and Limitations
- The author does not explain who wrad is or whether the NAI / NovelAI in the name implies training images were generated by NovelAI; it may involve imitating an artist's style.
- The sampler is not specified by the author; the license does not allow commercial use of generated images or derivatives, only site rental and rental; version notes are empty.
Trigger Word: @wrad
[S] Group Photo Control — Headcount, Positions, Height & Build per Person · MOSS
LoRA · Ported · Author 858654128784 · Published 11-22 11:53 · Downloads 118
Group photo control LoRA: Precisely specify the number of people, rows, and positions in a group photo via prompts, as well as each person's height, build, skin tone, hairstyle, clothing, and shoes from left to right; a MOSS port of the Krea 2 version released on the same day.
You can use the same template to write for few people, medium headcount, and close to 20 people to verify if the generated headcount and positions match. (Editor upgraded to S: novel capability, copyable prompts, permissive license; few visible examples are the main deduction.)
Images

Reasons for Recommendation
- New capability: In multi-person group photos, issues like extra/missing people, swapped positions, or everyone looking identical are old problems in text-to-image generation; the author claims it generates exactly the specified number of people and rows, with heights differentiated by centimeters (tallest and shortest fixed), builds categorized as slim / athletic / muscular / average / stocky / heavyset, allowing people of different builds to stand in the same row.
- Ported: The Krea 2 version and MOSS version were listed simultaneously, with 369 total downloads for the model page; the MOSS version had 118 downloads within 9 hours of release.
- Training data consists of 150 casual smartphone group photos taken in daylight, each with 2 to 20 people, in scenes like campsites, courts, stairs, beaches, dorms, and office building entrances; poses were intentionally not trained (only standing, sitting, squatting), and the author claims a separate group photo pose LoRA can be stacked with it later.
- Serves text-to-image generation for team photos, class/club group shots, and multi-character positioning compositions; the prompt template is very detailed and can be copied and modified.
- Permissive license: Allows image generation, rental, sale, and derivatives, with no attribution required.
How to Use
- Prompts start with "groupphotocontrol,"; the first sentence specifies total headcount, gender breakdown, number of rows, and location, followed by "Left to right in each row." then describe row by row, person by person.
- For each person, write in fixed order: ethnicity, exact age, height (in cm, marking shortest / tallest), build, skin tone, hairstyle, expression, pose (standing/sitting/squatting), clothing, shoes (write "shoes hidden" if feet are out of frame); optionally end with "Soft daylight, casual phone photo".
- LoRA strength 1.0; the author did not register a trigger word, use "groupphotocontrol" from the model card; steps, CFG, and sampler are not specified by the author.
Risks and Limitations
- Only 3 visible examples (another 3 sensitive-level examples hidden), which do not cover the 2 to 20 person range claimed by the author; you must verify if headcount remains accurate with larger groups.
- Total headcount, row count, and lines of per-person descriptions must match; do not write about flash or someone holding a phone, or a light or hand holding a phone may appear in the image; do not expect complex poses.
- Differences between the MOSS version and the Krea 2 version are not documented, and version notes are empty.
Recommended (8)
[A] HaDeS MOSS Suite Workflows · Merge, Save v0.9.4
Workflow · Version Update · Author _HaDeS_ · Published 11-22 20:28 · Downloads 17
MOSS model merge workflow: Mix MOSS SFT and Base models layer by layer, use telemetry nodes to check merge results, and save the merged model along with its recipe as your own model file.
(Editor upgraded to A: Auto-rating was suppressed to B due to incomplete scan; content is a merge tool with clear nodes and dependencies, but targets advanced users and relies on the author's node package.)
Images

Reasons for Recommendation
- Version update: Each version on this page is a different workflow; this version is the "Merge, Save, Telemetry" workflow accompanying the author's sixth video tutorial, following Metadata v0.9.3; the workflow JSON is not new weights.
- Node documentation is complete: The MOSS Model Block Merger's sliders control 6 layers each (0 retains SFT, 1 takes Base), with the last three sliders controlling text embeddings, time modulation, and input/output layers; recipe_override allows per-layer specification; also includes Model Saver, Metadata Reader, and a multi-model merger supporting TIES, DARE, Task Arithmetic, Model Stock, and SLERP (muted by default).
- Serves those wanting a middle ground between SFT's polished look and Base's diversity, or creating custom MOSS fusion base models; the author included self-tests, claiming merging only partially mitigates SFT's tendency toward East Asian faces when no person is specified, and specifying the person in prompts is most reliable.
How to Use
- Install HaDeS MOSS Suite v0.9.4 or newer (search HaDeS MOSS in ComfyUI Manager, or update from the author's GitHub), prepare MOSS-Image SFT and Base, the author-specified int8 text encoder, and FLUX.2 VAE (Hugging Face links provided in the model card).
- Built-in test image: seed 1013, 25 steps, CFG 2.5, er_sde / simple; the author claims the recipe in the video (18-35 layers mixed with half Base) is for demonstration only and must be tested by you; merged files are saved to output/diffusion_models and must be moved to models/diffusion_models to load.
Risks and Limitations
- Security scan status shows incomplete (file is just a 16 KB JSON config); relies on the author's own node package, and this tool was not run.
- The author claims the saved bf16 merged file is about 10 GB and int8 ConvRot is about 5.4 GB, resulting in large disk usage; speed test data comes from the author's RTX 5090.
- Update notes only contain the title and video link; the author claims old workflows will be gradually removed and replaced with new versions.
[A] Grimm's realistic painting · Variant 2-MOSS
LoRA · Ported · Author Grimm0 · Published 11-22 20:44 · Downloads 15
Realistic oil painting style LoRA: Heavy texture of classical oil paintings, expressive brushstrokes, and bright color transitions, between realism and impressionism, focusing on portraits and figure studies; a MOSS port of a multi-base model series.
(Editor downgraded to A: Art style direction is clear and examples are sufficient, but MOSS-specific parameters only include strength, and version differences are not explained.)
Images

Reasons for Recommendation
- Ported: The same page already has Krea 2, Anima, Flux.2 Klein 9B-base, ZImageTurbo, Illustrious, and Flux.1 D versions, with 4481 total downloads for the model page; the MOSS version filename indicates round 14.
- Serves text-to-image generation for portraits, figure studies, and romantic compositions, useful when wanting to render figures with an "oil painting brushstroke + soft transition" painterly feel; the author claims it balances realism and painterly quality.
- Visible 10 examples (another 2 sensitive-level examples hidden); trigger word follows the series' consistent GrimmsPaint.
How to Use
- Trigger word GrimmsPaint; strength 0.7 ~ 1 (general values given at the top of the model card).
- MOSS steps, CFG, sampler, and whether a turbo LoRA is needed are not specified by the author.
Risks and Limitations
- The version is named Variant 2, while the Krea 2 variant 1 on the same page is a different training set with a different trigger word; which dataset this MOSS version uses and how it differs from other base model versions is not documented, and version notes are empty.
- The model card is generic promotional copy, with no MOSS-specific parameters other than strength; released less than 1 hours ago, the 15 downloads for this version only represent early interest.
- License does not allow derivatives, and commercial use is limited to site rental.
Trigger Word: GrimmsPaint
[A] AG: Crocell - Painterly Backgrounds · MOSS-Image
LoRA · Ported · Author Azriaelis · Published 11-22 09:10 · Downloads 38
Crocell-style painterly background LoRA: A scene/background painting style with distinct brushstrokes in digital illustration (author tags: painterly, backgrounds, brushwork, artist style); MOSS port for Krea 2 Turbo and Anima versions.
(Editor downgraded to A: MOSS parameters are complete, but the trigger words emphasized by the author are missing in this version, and there is no textual description of the art style)
Image

Reasons for Recommendation
- Port: The same page already has Anima and Krea 2 Turbo versions, with the entire model page having 224 downloads; the author provided a separate set of settings for each of the three base models, and the MOSS set was provided individually, not copied over
- Version number 3469229 is earlier than other entries in this issue because it was cataloged in the same batch as the Krea 2 Turbo and Anima versions; the MOSS version was not previously public, its release time falls within this issue's window and was not included in the previous issue, making it a new version released for the first time in this issue
- Serves text-to-image generation for illustration backgrounds, scene concept art, and environmental images with a brushstroke feel; 17 visible examples (another 3 sensitive-level images are hidden), 12 hours after release with 38 downloads and 10 likes
How to Use
- MOSS settings: Base model + 8 step turbo LoRA, sampler euler, scheduler simple, LoRA strength 0.8-1, steps 8-16, CFG 1-3
- The author states that using trigger words significantly enhances the effect, but the MOSS version does not register trigger words; the trigger words for the Anima and Krea 2 Turbo versions on the same page are CR0C3LLX, and it is not stated whether the MOSS version uses the same ones, so you can try including them first
Risks and Limitations
- The automatically extracted parameters are misaligned: 0.6-0.8, 25 steps, CFG 4 are for Anima, and 0.8-1.4, 8-10 steps, CFG 1 are for Krea 2 Turbo; do not apply these to MOSS
- The MOSS version requires a 8 step turbo LoRA, but the author did not specify which one; the model card has no textual description of the art style, so you can only judge from the examples
- The name and artist style tags point to a specific art style, but the source and licensing are not stated by the author; the license does not allow derivatives; the version notes are empty
[A] Grimm's Fantasy · v5.5-MOSS
LoRA · Port · Author Grimm0 · Released 11-22 09:57 · Downloads 50
Fantasy + sci-fi concept illustration style LoRA: The author states it is aimed at fantasy characters such as warriors, elves, angels, and demons, surreal landscapes, and futuristic designs like robots and cyborgs; MOSS port for Krea 2, Anima, and SDXL versions.
(Maintained as A: Examples and attention are sufficient, but MOSS usage only lists trigger words)
Image

Reasons for Recommendation
- Port: The same page already has Krea 2 (including v5), Anima v5, and an SDXL test version, with the entire model page having 1006 downloads; the MOSS version is named v5.5, and the filename indicates the 14th round
- Serves text-to-image generation for concept design, world-building, and fantasy/sci-fi characters and scenes; the same author's Grimm's realistic painting leans towards realistic oil painting, while this one leans towards fantasy themes
- 11 visible examples (another 1 sensitive-level images are hidden), 11 hours after release with 50 downloads and 12 likes
How to Use
- Trigger word: grimmsfantasy
- The author did not provide MOSS strength, steps, CFG, or sampler; only the SDXL test version on the same page lists strength, which is for SDXL and cannot be directly applied to MOSS
Risks and Limitations
- The model card is thematic promotion and does not describe specific characteristics of the art style (lines, color palette, texture); judgment relies mainly on examples
- The differences between v5.5 and Krea 2 v5 are not stated, and the version notes are empty; the old version referenced by the model card is on the civitai.red mirror site, not on this page
- The license does not allow derivatives, and commercial use only includes on-site rental
Trigger Words: grimmsfantasy
[A] Dizzle's MIlky Way · v1
LoRA · New Model · Author Maridizzle · Released 11-22 18:57 · Downloads 18
Galaxy-themed concept LoRA: Composes people, creatures, and even staircases and desks into thousands of sparkling star points, connected by faint glowing lines, like outlines drawn with constellations; one of Dizzle's space series.
Image

Reasons for Recommendation
- New model: Unlike the same author's nebula version (which dyes the image with colored gas), this version makes the subject itself "composed of star points"; the author states that hair is stretched into glowing strands, wings become galaxies, and the ground reflects like black glass
- One of the few in the same series that provides "magic words": adding made out of swirling stars yields the best results; serves fantasy character, constellation/astrology-themed illustrations and posters
- 4 visible examples (another 3 sensitive-level images are hidden); free, license allows commercial use of outputs and derivatives, no attribution required
How to Use
- Trigger word: dz_milky, strength 1.0; add made out of swirling stars in the prompt
- The author did not provide steps, CFG, or sampler
Risks and Limitations
- Released about 2 hours ago, with 18 downloads and no likes yet; the model card is promotional copy with no training or parameter details
- One of the mass-produced themed LoRAs by the same author in the same batch; stability on subjects other than those shown in the examples is unknown
Trigger Words: dz_milky
[A] WILD AND WET [MOSS] · v1.0
LoRA · New Model · Author Franklin_Sports · Released 11-22 13:35 · Downloads 90
Retro print style LoRA for American sports and fashion magazines and advertisements from the 1989–1994s: The author trained on images scanned from their own old magazine clippings, mainly swimsuit posters, reproducing the texture and atmosphere of 90s magazine photos.
Image

Reasons for Recommendation
- New model: The training set consists of 500 magazine images scanned by the author, with network dim/alpha 32/16 and unet lr 0.0001 specified; the author states it can reproduce the appearance of swimsuits and people from that era, but cannot reproduce specific individuals
- Serves text-to-image generation for retro portraits, 90s advertisement and magazine interior page styles; besides the main trigger word, the author registered a set of era/print texture phrases that can be directly concatenated
- 7 hours after release with 90 downloads and 16 likes, making it the most downloaded in this group (only indicates attention)
How to Use
- Main trigger word: WAW90; add A vintage 1990s magazine pinup photo, authentic 90s sports magazine spread style, vintage print texture, printed on glossy paper, etc., for era and print descriptions
- Can continue with analog photography, shot on 35mm film, authentic film grain, soft focus background and faded color grading, nostalgic atmosphere.; the author did not provide strength, steps, CFG, or sampler
Risks and Limitations
- Only 3 visible examples, with another 17 sensitive-level images hidden; the training content is mainly swimsuit posters, showing a clear bias
- The training images are scanned commercial magazines and advertisements, and the author did not state copyright; the license does not allow commercial use of outputs, only including on-site rental and rental; it is not stated whether MOSS Base or SFT training was used
Trigger Words: WAW90, A vintage 1990s magazine pinup photo, authentic 90s sports magazine spread style, vintage print texture, printed on glossy paper, analog photography, shot on 35mm film, authentic film grain, soft focus background, faded color grading, nostalgic atmosphere.
[A] MOSS Skin Retouching Slider · v1.0
LoRA · New Model · Author baowy · Released 11-22 16:32 · Downloads 82
MOSS skin retouching slider LoRA, no trigger word needed: The higher the strength, the lighter the freckles and spots, the less redness, the fewer pores and fine lines, and the skin becomes more matte and whiter.
Running the same portrait through four levels: 0 / +3 / +5 / +7, allows you to directly see the boundaries of the author's tier table and where the face shape starts to change. (Editor upgraded from B to A: No visible examples, but the purpose, strength tiers, side effects, and prohibited scenarios are clearly written, so it can be used as is)
Image

Reasons for Recommendation
- New model: A utility LoRA that does not change the person or composition, only adjusts skin texture, used for realistic portraits, avatars, and portrait-style text-to-image generation, effectively doing a skin retouch pass during output
- The author provided a tier table: +3 light retouch, preserving believable texture; +5 natural retouch; +7 close to commercial retouch; +9 extreme, face shape starts to slim down; recommended range is +3 to +5
- The description lists side effects and limitations: It cannot override explicit shine in the prompt, spots only become lighter but do not disappear, skin tone becomes whiter; in scenes that should be wet, it will also wipe away water droplets
- 5 hours after release with 82 downloads and 9 likes (only indicates attention); the license is permissive, allowing sales and derivatives, no attribution required
How to Use
- No trigger word needed; strength +3 to +5, do not exceed +5 if consistency for the same person is required
- The author suggests pairing with a 8 step DMD acceleration LoRA, strength 1.0; the specific DMD file, CFG, and sampler were not provided by the author
- The more specific the prompt, the more stable: The author states that +7 is stable for avatar close-ups, faces filling the frame, and when background and lighting positions are specified; if the background is vague or the scene is complex, strengths above +5 may alter clothing and composition
Risks and Limitations
- Visible examples are 0: All 3 examples are listed as sensitive-level hidden, so before-and-after retouching comparisons are not visible on the website
- The author states that spot reduction is 15
25%, texture reduction is about 30%, and skin tone brightening is 810%, which are the author's own tests, with the testing method not stated - Strengths above +7 narrow the jawline and refine facial features, making the same person look younger; the author advises against using it for swimming, showering, or just-out-of-water scenes
- The card image is an AI-generated diagram, not the actual effect of the model; the author has no public non-sensitive examples
[A] Sequence7's Simple MOSS Multi-LoRA Workflow · v1.2
Workflow · Version Update · Author Sequence7 · Released 11-22 13:11 · Downloads 33
Sequence7's MOSS text-to-image ComfyUI workflow v1.2: On top of the minimalist "positive/negative prompts + multi-LoRA loading" flow, it adds a toggleable upscaler + second-pass sampling refinement, as well as face and hand repair.
(Editor upgraded to A: The update content is clearly written, new features are disabled by default and do not affect the original flow, and the model card also includes MOSS prompt experience; the reason for automatic downgrade is that the scan was incomplete, which has been noted in the risks)
Image

Reasons for Recommendation
- Version update: Compared to v1.1, it adds an optional upscaler and second-pass KSampler refinement, optional face/hand detailer, and the preview image is generated after the first sampling pass; all new features are disabled by default, the detailer preset is mild, and if not enabled, it is the same as the old workflow
- Retains the SkipLayerGuidanceDiT node added in v1.1 (skipping layer 22, the author states OpenMOSS also suggests trying layers 24 and 26), which the author claims reduces limb distortion
- The model card includes a set of MOSS prompt experience: Canvas aspect ratio should match the character pose, text to be rendered should be placed in straight double quotes, short prompts vary significantly with seed, switch to the SFT version if outputs diverge; the entire model page has 891 downloads and 29 likes
How to Use
- The default configuration is MOSS SFT INT8 quantization + DMD 8 step turbo LoRA, but it can also be changed to a full-weight model or a turbo model with built-in acceleration; you need to update ComfyUI to the latest nightly (the Load CLIP node must support the moss_image type) and install the Lora Manager node
- Do not enable a second sampling pass when using a turbo model or turbo LoRA; the author states this will cause color burning and loss of detail; a second pass is only recommended for full-weight models without an acceleration LoRA
Risks and Limitations
- The file security scan status is Error / Success and has not fully passed; check the JSON and the nodes it requires to be installed before importing
- The author did not specify which plugin the newly added face/hand detailer depends on; the workflow does not use new weights and does not inherently bring a new art style
- 2 visible examples are shown, while 12 sensitive-level examples are hidden
Worth Noting (19)
[B] Lilith Calah - Dreamkeepers · Moss
LoRA · Port · Author _mammoh · Published 11-22 15:40 · Downloads 16
A character LoRA for Lilith Calah, a green-haired cat-like character from the manga "Dreamkeepers," including six sets of clothing tags organized by manga volume; ported from NoobAI, Illustrious, Anima, Krea 2, and Pony versions.
Images

【B】2B (Nier Automata) - MOSS · v1](https://civitai.com/models/3061345?modelVersionId=3484171)
LoRA · New Model · Author Only4u_Art · Published 11-22 18:58 · Downloads 36
A LoRA for the character 2B from NieR: Automata, trained directly on MOSS: learning her face, bob haircut with black headband, mole below the mouth, eyepatch, and black dress.
Images

[B] Munin Female detective ( Silver palace) - MOSS · v1
LoRA · New Model · Author Only4u_Art · Published 11-22 17:04 · Downloads 21
A character LoRA for the detective Munin from the unreleased game "Silver Palace": learning her face, single drill curl, and detective outfit with a black fedora.
Images

[B] Stained Glass: Umbrium · MOSS v3
LoRA · Port · Author umbrium2835 · Published 11-22 19:00 · Downloads 22
A stained glass body LoRA: transforms exposed body parts like arms and legs into gem-colored glass pieces separated by dark seams, creating a translucent, glowing stained-glass texture while keeping the character recognizable; a MOSS port of the Krea 2 and Illustrious versions.
Images

[B] ccroquette_ | Style · MOSS(test)
LoRA · Port · Author AkioAI · Published 11-22 18:01 · Downloads 36
A LoRA imitating the anime illustration style of artist ccroquette, marked by the author as a MOSS test version; ported from the Anima version.
Images

[B] Dizzle's NGC 246 Nebula · v1
LoRA · New Model · Author Maridizzle · Published 11-22 16:45 · Downloads 18
A concept LoRA themed around the NGC 246 Nebula: fills the frame with glowing nebula gases in electric pink, purple, cyan, and blue, along with fine star specks, making rocks, skin, fabric, and furniture look as if woven from starry skies; one of Dizzle's space series.
Images

[B] JMOZ Style for Moss · v1.0
LoRA · New Model · Author Razira · Published 11-22 15:37 · Downloads 31
A cartoon comic style LoRA by artist Jmoz: focusing on exaggerated facial expressions and body proportions; the author admits that style reproduction on MOSS is inferior to Krea 2, and flat colors require other LoRAs.
Images

[B] Namah Calah - Dreamkeepers · Moss
LoRA · Port · Author _mammoh · Published 11-22 16:11 · Downloads 17
A character LoRA for Namah Calah, a character with white short hair, long horns, and a demon tail from the manga "Dreamkeepers," including three sets of clothing tags and negative prompts; ported from NoobAI, Krea 2, Anima, Illustrious, and Pony versions.
Images

[B] Dizzle's Volcanic Glass · v1
LoRA · New Model · Author Maridizzle · Published 11-22 12:51 · Downloads 25
A volcanic glass material concept LoRA: transforms armor, skin, hair, rocks, and weapons into polished obsidian with lava-like orange-red veins crawling inside, emitting an ember glow from within; one of Dizzle's batch releases.
Images

[B] Himari Azuma / Yachiho Azuma · v1.0
LoRA · New Model · Author Hayakake · Published 11-22 18:24 · Downloads 18
Two anime characters, Azuma Himari and Azuma Yachiho, combined into a single LoRA, with appearance tags and "Anti-demon corps" uniform tags written separately; the filename contains "mato," suggesting it is from "Chained Soldier."
Images

[B] Translucent Anime Style - MOSS · v1
LoRA · New Model · Author martinffm · Published 11-22 18:36 · Downloads 26
A translucent glowing anime style LoRA: the author states the overall image glows, featuring neon lights and translucent colors, with the trigger word "translucent."
Images

[B] FFstyle Anima/Moss Heal · v1.1 Moss heal
LoRA · Version Update · Author FEEEEF · Published 11-22 18:30 · Downloads 52
The second MOSS version (v1.1 Moss heal) of the FFstyle animation style LoRA; the author only wrote "nai5 ran it themselves," meaning the training images were generated by the author using NovelAI.
Images

[B] MOSS-CG style · v1.0
LoRA · New Model · Author linke666 · Published 11-22 14:22 · Downloads 92
A CG semi-realistic style LoRA trained for MOSS; the model card contains only a brief style description and recommended strength.
Images

[B] boleromix(moss-image) · v1.10
Base Model / Checkpoint · Version Update · Author bolero537 · Published 11-22 13:09 · Downloads 151
The second version of the MOSS merged base model boleromix, which the author describes as a "casual merge"; all seen examples are anime-style illustrations.
Images

[B] TOKKYU Style · v1.0
LoRA · New Model · Author gameplay666666 · Published 11-22 15:14 · Downloads 63
A TOKKYU style LoRA with no description provided by the author; seen examples are cute anime girl illustrations in pink and purple tones, often featuring decorative layouts with collages, stickers, handwritten text, and grid paper.
Images

[B] Star Trek DS9 Type 2 Phaser · MOSS Image-SFT - v1
LoRA · Port · Author lcarspicturesdatabase · Published 11-22 18:25 · Downloads 16
A prop LoRA for the 2-type phaser (Type 2 phaser) from the Star Trek: Deep Space Nine era, featuring a black utility belt equipped with a tricorder and a phaser; MOSS port for Qwen 2.1 and Krea 2 Turbo.
Images

[B] Real Skin · v1.5
LoRA · New Model · Author Sala666 · Published 11-22 17:29 · Downloads 47
A MOSS realistic skin LoRA: makes characters generated by the base model look more like real photos, with the author claiming minimal impact on the base model's original capabilities; recommended to pair with a 8 step DMD2 acceleration LoRA for fast realistic image generation.
Images

[B] Jinnsart Style (MOSS-Image) · v1.0
LoRA · New Model · Author Alt4Pai · Published 11-22 13:32 · Downloads 24
A LoRA imitating the Jinnsart style: flat colors and thick black outlines in an American cartoon style, trained on MOSS Base (int8).
Images

[B] mizuki shigeru esque [test] · test
LoRA · New Model · Author AS193SE · Published 11-22 12:29 · Downloads 20
A test version style LoRA imitating the atmosphere of manga artist Mizuki Shigeru, with the author stating it only learns the "atmosphere" and does not include characters.
Images

Included Only (15)
- Judy Hopps - David Lillie style · Moss — LoRA: A character LoRA for Judy Hopps from Zootopia, rendered in the art style of Dreamkeepers author David Lillie; ported from Anima, Krea 2, Illustrious, NoobAI, and Pony versions.

- Cinderella ( Silver palace ) - MOSS · v1 — LoRA: A character LoRA for Cinderella from the same game Silver Palace: learns her face, spiked crown, skin cracks, and damaged white dress, outputting only in anime style.

- Ural girls\Lana · v1.0 MOSS — LoRA: A realistic portrait character LoRA named "Lana" (from the Ural girls series), which the author claims is an original character, allowing changes to clothing, lighting, and camera angles; a MOSS port of the Krea 2 version.

- A2 ( Nier Automata ) - MOSS · v1 — LoRA: A character LoRA for A2 from NieR: Automata: learns her face, long white hair covering one eye, mole below the mouth, and damaged black outfit; the short-haired style after cutting her hair is also included in the training set.

- GC ELF STYLE · moss step4750 — LoRA: A style LoRA named GC ELF STYLE, with "elf" in the trigger words, but the model card does not specify what art style it is or what it is used for; a MOSS port of the Krea 2 version.

- OC Keira · Moss — LoRA: A realistic portrait LoRA for the original character Keira: a young woman with pink undercut spiky short hair, blue-gray eyes, and an emo vibe.

- hameko · v1.0 (moss) — LoRA: A character LoRA for the female character hameko with black and white two-tone hair, a MOSS port of the Illustrious and Anima versions.

- megauber-Jenn · MOSS Image — LoRA: A LoRA for an AI-generated virtual character, a young blonde woman, used to reproduce the same character across different scenes.

- Artist: kanzarin | draft | Moss · v1.0 — LoRA: A draft version LoRA in the art style of artist kanzarin, which the author merged from epoch 24 and epoch 40 at a ratio of 0.7:1.0.

- Taurvi LaDonna & Zarich LaDonna- Langrisser Millennium · v1 Zarich<MOSS> — LoRA: A character LoRA for Zarich LaDonna (green hair) from Langrisser Millennium.

- 123moss art style lora · v1.0 — LoRA: An art style LoRA that the author describes as "just thrown together," with no specification of the art style; the filename is miejianguo_s1440.

- MOSS Image Korean and Chinese Drama Style · v1.0 — LoRA: A LoRA for Korean and Chinese drama still styles, available for paid download; the examples seen are realistic still-style images of East Asian women in everyday scenes such as bedrooms, buses, and cafes.

- LeonidAfremov Style · v1 — LoRA: An art style LoRA named after painter Leonid Afremov, with the trigger phrase "An LeonidAfremov Style painting of"; the model card contains no description other than the title.

- Stylized Anime Aura - for Moss · v1.0 — LoRA: A modern anime illustration style LoRA: the author claims it brings clean line art, detailed clothing textures, and a cinematic atmosphere with warm soft lighting; available for paid download.

- bracedfaced exotic woman (sorry bout the title but civit) STCK · v1.0 — LoRA: A paid character concept LoRA, which the author states is used to generate exotic-style women wearing braces.

Data: Civitai public API (base model MOSS Image), inclusion window 11-22 08:31 → 11-22 21:01 (Beijing Time). Recommendations have been reviewed by editors; download counts only reflect attention.