MOSS Image Evening Report 2026-11-01 | Top Pick: MOSS Anime ↔ Realistic Slider

Zhenzhen's AI Radar · MOSS Image · 2026-11-01 Evening Report

This issue features 51 new versions: Strongly Recommended 1 · Recommended 10 · Worth Watching 18 · Listed Only 22

Editor's Note This issue reviews 52 new MOSS versions. The top pick is the Anime ↔ Realistic Style Slider: negative weights lean toward Japanese anime illustrations, positive weights lean toward realistic photography; no trigger words are needed, and sampling parameters including SLG are fully provided, maintaining an S rating. Other highlights include scene LoRAs with detailed prompt writing for styles such as single-panel comics with readable captions (mimicking commercial comic strips, downgraded from S to A), Chinese dreamcore scenes, rainy neon cities, hard sleeper train compartments, and home DV frame-grab textures, as well as a workflow for generating multi-view character sheets via text. Several ported or style LoRAs in the automatic S/A list lack MOSS-specific parameters or descriptions and are downgraded to B; standard character/persona LoRAs and entries with only one or two sentences of description are downgraded to C. The two files named V2.2 under Moody Vivid Mix are different precision versions (free INT8 / FP8 and paid BF16), so they are not considered duplicates and both are retained.

Strongly Recommended (1)

[S] MOSS Anime ↔ Realistic Slider · v1.0

LoRA · New Model · Author Lumir · Published 11-01 20:11 · Downloads 23

A style slider LoRA for MOSS-Image-SFT: negative weights lean toward modern Japanese anime illustrations, positive weights lean toward realistic photography, and 0 represents the base model as-is; no trigger words are required.

Maintains S rating: The capability is clearly explained in one sentence, works in both positive and negative directions, sampling parameters including SLG are fully provided, there are plenty of examples, and it has zero barrier to entry.

Images

Reasons for Recommendation

• New model: Solves the need to continuously adjust between anime illustration and realistic photography using the same prompt, without switching base models or finding separate style LoRAs.
• Parameters are complete and actionable: Start adjusting from -1.0 or +1.0; use er_sde + simple, 25 steps, CFG 3, 1024 × 1024; for SLG, select uncond skip, layer 26 (counting from 0), range 0.0–1.0; applies only to the model, CLIP strength 0, leave negative prompt empty.
• 10 visible examples, some of which are wide collage sizes; based on dimensions, they appear to be comparison images with different weights (not verified by opening); free, license allows merging; published less than 1 hours ago with 23 downloads (indicating attention only).

How to Use

• Load moss_anime_realistic.safetensors (16 MB · FP32), use a LoRA node that loads only the model (CLIP strength 0), no trigger words; weight -1.0 leans toward anime, +1.0 leans toward photography, adjust as needed.
• er_sde / simple, 25 steps, CFG 3, 1024 × 1024, SLG uncond skip layer 26, range 0.0–1.0, leave negative prompt empty.

Risks and Limitations

• The author only provided starting points of ±1.0; performance and limits at higher weights are not specified.
• The author's description is specific to MOSS-Image-SFT; performance on the Base model is not specified.
• License does not allow commercial use; publication time is very short, with few downloads and feedback; this entry has not been tested.

Recommended (10)

[A] Vinne Style | MOSS-Image · v1.1

LoRA · Version Update · Author deitychaser · Published 11-01 15:06 · Downloads 43

The second MOSS version of the 'Vinne' style LoRA: a thick paint, gritty textured, low-saturation illustration style (tags include retro, cyber, gothic); this version was retrained by a different trainer.

Downgraded from S to A: Parameters are complete, examples are plentiful, and the license is permissive, but the update notes are nearly empty, the style description is limited to one optional prompt sentence, and the training data source is unclear.

Images

Reasons for Recommendation

• Version update: Compared to v1.0, v1.1 was retrained by collaborator NerissG using his experimental training settings; the author claims this version is the 'winner' in comparisons; update notes only mention trigger words, not changes to data or training parameters.
• Serves users wanting thick paint, coarse grain, and low-saturation illustrations in MOSS; optional style prompts on the v1.0 card (painterly, heavy gritty texture, muted color palette) help understand the style direction.
• Parameters are actionable: Trigger words, strength, sampler, scheduler, steps, and CFG are all provided, along with two recipes using DMD2 acceleration LoRAs; 14 visible examples; free, fully open license; this version has 43 downloads, cumulative model downloads 336 (indicating attention only).

How to Use

• Load Vinne_neriss_moss.safetensors (36 MB), write '@vinne' or 'Art by @vinne' in the prompt, strength 0.8-1.0.
• Standard: dpmpp_sde + Simple, 16 steps (author claims equivalent to 32 steps), CFG 4.
• Accelerated: Use euler + simple with DMD2 Turbo LoRA, 8 steps, CFG 1; or set DMD2 LoRA to 0.5, use euler for 16 steps, CFG 2-3.

Risks and Limitations

• Update notes contain only one line of trigger words; differences between v1.1 and v1.0 can only be judged from examples; the 15 steps, CFG 3, Euler+Simple, and Layerskip node suggestions for v1.0 do not apply to v1.1.
• The style is named after a specific artist, but the training data source is not specified; it may be imitating a particular illustrator.
• The author did not provide the specific file for the DMD2 Turbo LoRA.
• 6 sensitive-level examples are hidden; this entry has not been tested.

Trigger Words: @vinne, Art by @vinne

[A] Farside Style · v2.0 MOSS

LoRA · Port · Author colinu · Published 11-01 13:49 · Downloads 31

The MOSS version of the 'Farside Style' single-panel satirical comic LoRA: generates black-and-white single-panel joke comics with a readable line of English caption at the bottom (the punchline is in the caption).

Downgraded from S to A: The capability for single-panel comics with readable captions is specific, and both prompt templates and complete/accelerated parameter sets are provided; however, the style explicitly mimics a published commercial comic strip, requiring self-assessment of copyright risks before commercial use, and 5 sensitive-level examples are hidden, so it is not a top pick.

Images

Reasons for Recommendation

• Port: Previously available as a Z-Image Turbo version (v1.0 ZiT, empty version notes); the MOSS version was retrained on MOSS-Image SFT, and the author claims this is the first truly usable version, capable of generating single-panel comics with clear, readable captions.
• Serves users needing humorous illustrations, newspaper-style comics, or meme scenarios requiring 'one image + one joke'; the author provides a complete prompt format: start with the trigger word, describe a clear scene, and write the punchline in the caption.
• Parameters are complete and actionable: Strength 1.0, full quality 50 steps, CFG 3, euler / sgm_uniform; the 4 step Turbo recipe used for example images is also provided; 11 visible examples, most with prompts; free; this version has 31 downloads in about 7 hours since publication, cumulative model downloads 208 (indicating attention only).

How to Use

• Load farside-moss.safetensors (138 MB · BF16), start the prompt with 'farside', e.g., 'farside a black and white cartoon of ... At the bottom of the image is the caption "..."'; keep captions short and use common words.
• Full quality: Strength 1.0, 50 steps, CFG 3, euler / sgm_uniform.
• Fast generation: Stack MOSS SFT 4 step Turbo LoRA (Kijai rank-72), 4 steps, CFG 1, euler / simple; the author notes that Turbo occasionally misspells rare words, so try multiple seeds.

Risks and Limitations

• The author states that images look flat without captions; effectiveness depends on the caption.
• The style explicitly mimics the published commercial comic strip 'The Far Side'; assess copyright and trademark risks before commercial use; license does not allow merging.
• Rare words may be misspelled with the Turbo recipe, requiring seed filtering.
• 5 sensitive-level examples are hidden; this entry has not been tested.

Trigger Word: farside

[A] AG: Beleth - Niji 7 Boys · MOSS-Image

LoRA · Port · Author Azriaelis · Published 11-01 11:00 · Downloads 29

The MOSS version of 'Beleth' from the 'Ars Goetia' demon series style LoRAs: labeled as Niji 7 style, sketchy, semi-realistic male character illustrations; the author claims the style is significantly stronger with trigger words.

Downgraded from S to A: MOSS section parameters are largely provided and examples are sufficient, but step counts are incorrect, it relies on an unspecified Turbo LoRA, and training data includes AI-generated images.

Images

Reasons for Recommendation

• Port: Previously available as Krea 2 and Anima versions; the MOSS version has empty release notes, but the model card lists MOSS section parameters separately (MOSS-Image Base stacked with a 8 step Turbo LoRA), which differ from the parameters in the Anima and Krea 2 sections.
• Serves users wanting male characters and sketch-textured semi-realistic illustrations in MOSS; the author describes this series as casual small experiments, with varying quality across entries.
• 12 visible examples (with prompts); free; this version has 29 downloads, cumulative model downloads 402 (indicating attention only).

How to Use

• Load BELETH-MOSS-V1 (1).safetensors (246 MB · BF16), write the trigger word 'B3L3TH77' in the prompt.
• Stack a 8 step Turbo LoRA on MOSS-Image Base: euler + simple, strength 0.8-1, CFG 1-3; the step count is written as '-816' on the card, which the author did not clarify; the specific Turbo LoRA to use was not provided by the author.

Risks and Limitations

• Training images come from mixed sources including Midjourney, NijiJourney, Tumblr, etc., containing other AI-generated images and web images of unknown origin.
• The automatically scraped 25 steps, CFG 4, strength 0.6-0.8 are Anima section parameters, and 8-10 steps, strength 0.8-1.4 are Krea 2 Turbo section parameters; neither applies to MOSS.
• The MOSS section step count is written as '-816', possibly a typo for 8–16 steps; the required 8 step Turbo LoRA is unspecified.
• License does not allow merging; 4 sensitive-level examples are hidden; this entry has not been tested.

Trigger Word: B3L3TH77

【A】MOSS-Image Character Sheet Designer for ComfyUI · v1.0

Workflow · New Model · Author ukr8b3g201 · Published 11-01 18:58 · Downloads 39

MOSS Character Sheet (Multi-View Turnaround) ComfyUI Workflow: Describe the character's appearance and clothing via text to automatically or manually layout a character reference sheet with up to 13 views; alternatively, let the MOSS text encoder describe an image first, then generate from text.

Downgraded from S to A: Multi-view character sheets are a clear workflow capability, and starting parameters are provided; however, it requires installing custom nodes and accompanying models, only 2 visible examples are available, and image input is merely converted to text, so expectations should be adjusted accordingly.

Images

Reasons for Recommendation

• New Model: The author previously created H3 and Qwen-Image 2.1 version character sheet workflows. The MOSS version has a similar interface and most features (visual view selection, turnaround sheets, clothing details, part-specific prompts, styles, automatic/manual layout); however, the author specifically notes that the MOSS version is pure text-to-image, where input images are only converted into text descriptions and not used as reference images.
• Suitable for users involved in character design, early-stage concept art, or creating character reference sheets for video or comics; the author provides prompt writing guidelines: describe only one character's face, hairstyle, body type, clothing, color scheme, and accessories, without mentioning backgrounds or other characters.
• Free, with fully open site licensing; 2 visible examples; approximately 2 hours since release with 39 downloads (indicating popularity only).

How to Use

• Git clone the author's MOSS-Image-Character-Sheet-Designer node repository into ComfyUI's custom_nodes directory, restart, and import the workflow from mossImageCharacterSheet_v10.zip (6 KB); requires a newer version of ComfyUI that supports native MOSS-Image and subgraphs, with the CLIP loader type set to moss_image.
• Accompanying models: MOSS-Image SFT base model (or the MOSS_LUC listed by the author), text encoder for MOSS, FLUX.2 small decoder VAE, and lodestones' 8 step DMD acceleration LoRA; download links are available in the full model card and GitHub README.
• Author's starting settings: Euler / Simple, 8 steps, CFG 2-3; the author states these are just starting points and not guaranteed to be optimal; if you want to start from an image, load the image and enable enable_prompt_enhancer.

Risks and Limitations

• Not image-to-image or reference image control: When starting from an image, the face, clothing, color scheme, accessories, and unseen angles may change; the author explicitly states not to expect faithful reproduction of the original image.
• The author does not guarantee correct perspective, character consistency, or separation of clothing parts.
• Requires installing the author's custom node repository and multiple accompanying models; DMD LoRA strength (author's starting value 0.7-0.8) and sampling parameters are only documented in the full model card.
• 5 sensitive-level examples are hidden, exceeding the number of visible examples; the author has not provided update notes (first version); this entry has not been tested.

【A】Dreamcore — The School Corridor That Was Never This Long | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 11:48 · Downloads 48

Chinese Dreamcore scene LoRA: Depicts childhood locations in China from the 90s to the 2000s, such as school corridors with blue doors and green wainscoting, terrazzo floors, old residential compounds with floral bedsheets hanging out to dry, bathhouses, video rental shops, hot water rooms, and waiting halls, featuring warm hazy light, faded pastel colors, and film grain, while including one element that feels "off" in the scene.

Maintained at A: The subject matter is new, the prompt recipe is detailed enough to follow, and examples are sufficient; however, it lacks strength and sampling parameters, preventing an S rating.

Images

Reasons for Recommendation

• New Model: Not just an art style, but a specific list of locations combined with the "familiar yet wrong" dreamcore technique (endless corridors, climbing frames with too many poles, doors too small to enter).
• The prompt recipe is actionable: Start with zhongshi_menghe, Chinese dreamcore, name the location and specific objects, add one incongruity (infinite repetition, incorrect proportions), and end with warm hazy daylight, faded pastel colors, film grain; the author notes that fewer people and emptier rooms make it feel more like a dream.
• 10 visible examples (with prompts); free, with fully open licensing; approximately 9 hours since release with 48 downloads (indicating popularity only).

How to Use

• Load zhongshi_menghe.safetensors (142 MB · FP16), start the prompt with "zhongshi_menghe, Chinese dreamcore," then describe the location, objects (terrazzo, green wainscot, enamel mugs, floral bedsheets) and one incongruity, ending with warm hazy daylight, faded pastel colors, film grain.
• If including people, describe their appearance, age (e.g., an adult man aged 23), position, and action; to make it look more like a photo, you can stack a realistic LoRA with a strength of approximately 0.3.
• The author has not provided the LoRA's strength, steps, CFG, or sampler.

Risks and Limitations

• The trigger word zhongshi_menghe is not registered in the version information, only mentioned in the description.
• Lacks strength and sampling parameters; you can only refer to the prompts accompanying the example images.
• 10 sensitive-level examples are hidden, equal in number to the visible examples.
• Training data sources are not specified; this entry has not been tested.

【A】Underground Parking — The Cameras Are Always On, Nobody's Watching | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 16:32 · Downloads 53

Underground parking scene concept LoRA: Depicts a garage with white-lined parking spaces, floor arrows, blue-trimmed pillars, light strips, and pipes; allows switching between eye-level camera angles or surveillance views from ceiling corners (with a line of surveillance text at the top).

Maintained at A: The scene and prompt writing methods are clearly explained and easy to follow, but there are no sampling parameters, and only 4 visible examples.

Images

Reasons for Recommendation

• New Model: Turns a specific scene into a controllable template; the author details scene elements and three camera angles (standing eye-level, ground-level low angle, ceiling corner surveillance view).
• Suitable for realistic scene generation, suspenseful narratives, or surveillance-style illustrations; the author provides a prompt order: camera angle, vehicles/pillars/lighting, character appearance/age/position/action/expression.
• All 4 visible examples include prompts; free, with fully open licensing; approximately 5 hours since release with 53 downloads (indicating popularity only).

How to Use

• Load undergroundparkingconcept.safetensors (142 MB), start the prompt with "undergroundparkingconcept," first describe the camera angle, then cars, pillars, floor arrows, and lights, and finally each character.
• For surveillance views, write out the top text completely, e.g., "CAM 12 B2 PARKING"; the author notes that omitting this will result in garbled text; write character ages in the format "an adult man aged 23."
• The author has not provided strength, steps, CFG, or sampler.

Risks and Limitations

• The trigger word is not registered in the version information, only mentioned in the description.
• No sampling parameters are provided.
• 13 sensitive-level examples are hidden, far exceeding the 4 visible examples.
• This entry has not been tested.

【A】Sleeper Train — Twenty Hours in a Hard Sleeper, You Need Something to Do | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 09:35 · Downloads 66

Chinese train hard sleeper carriage scene concept LoRA: Depicts blue upper and lower bunks, folded white quilts, small iron ladders and railings at the end of berths, small folding tables under windows, fields passing by outside, and narrow aisles with folding seats; features two lighting conditions: daytime and nighttime.

Maintained at A: Scene knowledge is specific, prompt writing methods are detailed, and visible examples exist; however, it lacks sampling parameters, and the main examples are hidden, preventing an S rating.

Images

Reasons for Recommendation

• New Model: Enables MOSS to recognize the specific furnishings and lighting of domestic train hard sleeper carriages (daytime window light, nighttime ceiling lights and corridor blue light); general base models typically only generate generic train carriages.
• Suitable for travel narratives, domestic life scenes, and realistic images of people in carriages; the author provides a prompt order: trigger word, first describe the camera angle (looking diagonally from the door or from the opposite lower berth), then bunks, bedding, ladders, tables, windows, and finally each person's appearance, age, berth location, action, and expression.
• The author also provides tips to avoid errors: describe the ladder as fixed to the upper bunk frame; 5 visible examples with prompts; free, with fully open licensing; approximately 11 hours since release with 66 downloads (indicating popularity only).

How to Use

• Load sleepertrainconcept.safetensors (142 MB · FP16), start the prompt with "sleepertrainconcept" (not registered in version information), and write in the order: camera angle → carriage furnishings → characters.
• The author has not provided strength, steps, CFG, or sampler.

Risks and Limitations

• 15 sensitive-level examples are hidden, three times the number of visible examples; part of the description specifically addresses adult content, noting that in narrow spaces, failing to specify where characters' knees and feet land will result in floating limbs and anatomical errors.
• The trigger word is only mentioned in the description and not registered.
• Strength and sampling parameters are not provided by the author; this entry has not been tested.

【A】Cyber Night — A City in the Rain, Lit by Neon | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 12:54 · Downloads 69

Post-rain city night photography style LoRA: Features puddles on the ground reflecting long, colorful streaks from neon and car lights, with cool blue distant buildings and warm orange-red nearby lights, fine rain in the air, and soft glows around every light source; the author emphasizes this is realistic cinematic color grading, not illustration.

Maintained at A: The style definition and prompt structure are among the clearest in this group, and examples are viewable; however, it lacks strength and sampling parameters.

Images

Reasons for Recommendation

• New Model: The author describes visual elements specifically (puddle reflections, warm/cool contrast, rain streaks, light glows), stating it is applicable to any night scene.
• Suitable for cyberpunk street scenes, rainy night portraits, and realistic cinematic city nights; the author provides a prompt order: first describe the camera angle (low angle shooting wet pavement, through convenience store glass, looking up from street level), then location, neon, and reflections, and finally each character.
• 5 visible examples, all with prompts; free, with fully open site licensing; approximately 8 hours since release with 69 downloads and 10 likes (indicating popularity only).

How to Use

• Load cybernightstyle.safetensors (142 MB · FP16), start the prompt with "cybernightstyle" (only mentioned in description, not registered), and write in the order: camera angle → scene → neon and reflections → character appearance, age, action, expression.
• Vertical composition works best; for signs, either specify the text or explicitly state there is no text, otherwise garbled characters will appear; the author has not provided strength, steps, CFG, or sampler.

Risks and Limitations

• No generation parameters are provided; strength must be determined manually.
• The trigger word is not registered in the version information.
• 12 sensitive-level examples are hidden, exceeding the number of visible examples; this entry has not been tested.

【A】Volunteer Recruitment — Who Signs Up for This? | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 17:04 · Downloads 37

Volunteer recruitment app interface concept LoRA: A vertical mobile interface featuring a large natural-light portrait photo at the top, with a cream-colored card below containing a dark green title, three lines of icons for meeting location/time/number of registrants, organizer avatars, and status badges; the author states it can generate interfaces in both Chinese and English.

Maintained at A: The text-based interface layout is described in detail and prompts are easy to write, but there are no sampling parameters, and only 4 visible examples.

Images

Reasons for Recommendation

• New model: Turns a fixed app interface layout into a template, with the author clearly defining every element and its position within the cards, emphasizing that photos are the main subject and the interface is secondary
• Suitable for scenarios such as app interface mockups, event posters, and product demo images; titles, locations, times, registration counts, and statuses (Approved / Pending / Full) can all be specified using quotation marks
• All 4 visible examples include prompts; free, with an open license; released about 4 hours ago with 37 downloads (indicating attention only)

How to Use

• Load volunteerrecruitmentconcept.safetensors (142 MB), start the prompt with "volunteerrecruitmentconcept,", and first describe the people in the photo (appearance, age such as "an adult man aged 23", volunteer vest, actions, scene)
• Then use quotation marks to specify the event title, location, time, number of registrants, and status, ending with "there is no other text"
• The author did not provide strength, steps, CFG, or sampler settings

Risks and Limitations

• The trigger word is not registered in the version information, only mentioned in the description
• No sampling parameters are provided; the stability of Chinese interface text is only briefly mentioned by the author
• 16 sensitive-level examples are hidden, far more than the 4 visible examples
• This entry has not been tested

[A] Home DV — The Tape That Should Have Stayed in the Drawer | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-01 10:40 · Downloads 157

A LoRA for the visual style of home DV tapes from the 2000s: 4:3 aspect ratio, soft details, noise in shadows, blown-out highlights, an overall warm yellow tone, and a white date watermark in the bottom right corner.

Maintaining A: The style characteristics and prompt structure are clearly explained, and the tips for avoiding pitfalls are practical; however, there are few visible examples and missing sampling parameters.

Images

Reasons for Recommendation

• New model: Creates the texture of "old home video frames"; indoors features flat lighting on nearby faces from the camera's built-in light, paired with warm table lamps and grainy brown shadows; outdoors features harsh direct sunlight and slightly overexposed skies
• Suitable for realistic images with nostalgic, pseudo-documentary, or suspenseful atmospheres; the author provides a prompt structure: trigger word, first describe the low camera angle and foreground objects, then describe the person's appearance, age, position, and actions, ending with the fixed phrase "home DV video still, soft detail, slight video noise, white date stamp at the lower right"
• The author provides tips to avoid pitfalls: do not write "camcorder" or "camera", otherwise the device itself will appear in the frame; for first-person shots, only describe the small part of the body visible in the frame; if the image is too soft, overlay a realistic LoRA with low weight; 3 visible examples with prompts; free, with an open license; released about 10 hours ago with 157 downloads (indicating attention only)

How to Use

• Load homedvconcept.safetensors (142 MB · FP16), start the prompt with "homedv2000," (not registered in the version information and different from the filename), and end with the author's provided "home DV video still ..." phrase
• The author did not provide strength, steps, CFG, or sampler settings; if the image is too soft, overlay a realistic LoRA, with the author only specifying "low" strength

Risks and Limitations

• Only 3 visible examples, with 16 sensitive-level examples hidden; the model name itself leans towards private recording themes
• The trigger word is only mentioned in the description and not registered
• The author did not provide strength or sampling parameters; this entry has not been tested

Worth Watching (18)

[B] MossStock · V1

LoRA · New Model · Author Adel_AI · Released 11-01 16:16 · Downloads 40

A MOSS-style LoRA named "MossStock", with the trigger word "Bradhamel art style"; the author did not describe the style in text, only providing the sampler and strength.

Images

[B] Sam_MOSS · Turbo v2.5

Base Model / Checkpoint · Version Update · Author toya_san · Released 11-01 15:09 · Downloads 129

A MOSS base model (Checkpoint) attempting to create a "Turbo" low-step image generation; the author claims it is usable with 8 steps or more; v2.5 aims for the cleanest and sharpest images, at the cost of slightly reduced prompt adherence.

Images

[B] Moody Vivid Mix - Moss Image (Free Download) · V2.2 25STEPS CFG3 BF16

Base Model / Checkpoint · Version Update · Author catlover1937 · Released 11-01 16:36 · Downloads 201

V2.2 of the "Moody Vivid Mix" realistic photography-oriented MOSS fusion base model: the author claims it focuses on high-contrast, high-saturation vivid color realistic photos, with diverse faces and good prompt adherence.

Images

[B] Saturday Morning Cartoon Style · MOSS

LoRA · Ported · Author vizsumit · Released 11-01 16:54 · Downloads 25

The MOSS version of the "Saturday Morning Cartoon" style LoRA: the author claims it aims to recreate the cartoon style of American TV Saturday morning children's animation blocks.

Images

[B] Limbus Company Identity style · MOSS v1.0

LoRA · Ported · Author Xegakup · Released 11-01 15:54 · Downloads 39

The MOSS version of the LoRA for the character portrait (Identity) style of the mobile game "Limbus Company"; the model card only contains a list of character prompts written for the Krea 2 version, with no description for the MOSS version.

Images

[B] Galena CAT - Galena Citron Anime Treasure [MOSS & ANI & IL] · MOSS

Base Model / Checkpoint · Ported · Author CitronLegacy · Released 11-01 11:37 · Downloads 105

The MOSS version of the anime fusion base model "Galena CAT": based on MOSS-Image SFT, incorporating the author's two unpublished MOSS LoRAs (MoonlightCAT, CuteRetroStyle), continuing the series' anime style with a focus on lips, colors, and depth-of-field blur.

Images

[B] Mrpopo's anime-moss-image · V2.0 further trained

LoRA · Version Update · Author Mrpopo714 · Released 11-01 10:00 · Downloads 63

The second version of a general anime style LoRA, trained longer on the same dataset using MOSS-Image Base; the author describes it as experimental and does not detail the style.

Images

[B] V67 Abstract anime · V1 MOSS

LoRA · Ported · Author angelomaiota · Released 11-01 17:18 · Downloads 15

The MOSS version of "V67 Abstract Anime": mixes anime characters with abstract expressionism, featuring splattered paint, fragmented compositions, neon high-contrast colors, and gray borders on the sides.

Images

[B] Service Arrival — "Your Worker Is On The Way" Card | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-01 17:10 · Downloads 45

A concept LoRA for generating screenshots of the "Your Worker Is On The Way" page in a home service app: a vertical interface with a service provider card at the top featuring a real photo, name, rating, and call button, followed by arrival time, a progress bar (Order Accepted → On the Way → Arrived), an order card, a mini-map, and a chat bubble; supports both Chinese and English interfaces.

Images

[B] Locker Room — 更衣室 | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-01 14:14 · Downloads 115

A locker room scene LoRA: rows of lockers on both sides, some doors open with clothes hanging, a long wooden bench in the middle with sports bags, towels, and water bottles, tiled floor, and mirrors, sinks, and showers at the far end.

Images

[B] Volumetric Glow Fantasy Style Illustration Aesthetics · v1.0

LoRA · Ported · Author Kinrol · Released 11-01 11:33 · Downloads 62

The MOSS version of a volumetric light fantasy illustration style LoRA: described by the author as featuring volumetric light, soft cinematic lighting, glowing materials, and soft focus diffusion, leaning towards the rendering quality of high-end gacha game character art.

Images

[B] Body Scanner — What the Airport Scanner Saw | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-01 17:07 · Downloads 53

A concept LoRA for an airport security full-body scanning scene: a realistic eye-level photo of a security checkpoint, featuring a silver dome scanner, yellow footprint floor mats, conveyor belt trays, and an operator's monitor displaying a dark blue interface with a light-colored human silhouette and a green CLEAR or amber warning box.

Images

[B] konya-karasue moss image · v1.0

LoRA · New Model · Author Eustoma2 · Released 11-01 17:33 · Downloads 40

A lighting LoRA: trigger word konya-karasue, used in conjunction with the lighting tags listed by the author (moonlight, window light, rim light, backlight, silhouette, chiaroscuro, volumetric light, Tyndall effect, low-key lighting, etc.) to control lighting; adding warm_lighting for warmer tones and cool_lighting for cooler tones.

Images

[B] Holi Festival of Colors Moss Image by Sarcastic TOFU · Holi_Festival_Moss_v1.0

LoRA · New Model · Author sarcastictofu · Released 11-01 11:11 · Downloads 33

A LoRA themed around the Indian Holi Festival: adds effects of colorful powder filling the air, colored smoke, and wet colored paint splashed on people.

Images

[B] Avant-Garde Style LoRA · v1.0 (MOSS Image)

LoRA · New Model · Author frutiemax · Published 11-01 10:02 · Downloads 28

The MOSS version of the Avant-Garde style LoRA; the author only wrote one sentence, "Generate avant-garde style models," with no further description.

Image

【B】AuthentiQ Niji Ink-Engraving Style [MOSS] · MOSS_v1.0

LoRA · New Model · Author AuthentiQ · Published 11-01 18:29 · Downloads 33

A semi-realistic style LoRA in the Niji V7 style; the author states it adds extra outlines and engraving effects; tags point to poster-style illustrations.

Image

【B】Gallery — Phone Photo Album Page, One Album One Story | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-01 15:07 · Downloads 67

A concept LoRA for phone photo album app interface screenshots: status bar, "< Albums / Select" top bar, album titles with dates and counts, a large image with colored borders and like/share/delete icons, and 2×2 thumbnails below.

Image

【B】MOSS-Image TestSampler and Merger for ComfyUI · v1.1

Other · Version Update · Author aueki4g467 · Published 11-01 14:27 · Downloads 6

A ComfyUI model development toolkit specifically for MOSS-Image: includes the TestSampler for testing, a Layer Merge node for merging by layer, and a converter to transform INT8 models into merged BF16; this version fixes the Merger's default settings and adds slider shortcuts.

Image

Listed Only (22)

• Kira Zima — Photorealistic LoRA [MOSS / Krea] · v1.0 MOSS — LoRA: The MOSS version of the realistic character LoRA "Kira": a female character with Slavic features, the author emphasizes soft realism, clean skin, facial consistency, and no plastic look.
• Clea (Expedition 33) Moss · v1.0 — LoRA: A MOSS character LoRA for Clea from the game Clair Obscur: Expedition 33: long curly auburn hair, bangs, black headband, grey-blue eyes; clothing (pinstripe shirt, suspenders, dark skirt) can be added as needed.
• Tattooed Babe Moss Image by Sarcastic TOFU · Tattooed_Babe_Moss_v1 — LoRA: A LoRA for female characters covered in tattoos; the MOSS version's model card directly reuses the text from the Krea 2 version.
• Sweaty Cutie Moss Image by Sarcastic TOFU · Sweaty-Cutie_Moss-v1 — LoRA: A concept LoRA that adds sweat, body oil sheen, or water droplet effects to female characters' skin; the author states it works for both indoor and outdoor scenes.
• Midnight/Blacknight (Arknights) · v1 — LoRA: A character LoRA for the Arknights operator Midnight (Blacknight): covers Elite 0, Elite 2 original skins, and two skins—Summer Flora FA641 and Branch Fragrance; complete copyable tags are provided for each outfit.
• south-indian-beauty - moss -sft · moss-sft — LoRA: A character appearance LoRA for "South Indian beauty," trained on MOSS-Image SFT using AI-Toolkit; the model card only lists the trigger word and training epochs.
• Evelyn Parker Cyberpunk2077 · v1.0 — LoRA: A character LoRA for Evelyn Parker from Cyberpunk 2077: blue bob cut, straight bangs, freckles; includes trigger phrases for gold fingertip accessories and two outfits.
• Kali Belladonna - RWBY - MOSS · V1 — LoRA: A character LoRA for Kali Belladonna from RWBY; appearance tags and default outfit (white kimono with black hakama, etc.) tags are provided separately; the author states this is their first time training a MOSS LoRA.
• [MOSS-Image] Classic Arcaea Game Style LoRA · alpha1 — LoRA: An anime-style LoRA imitating the classic art style of the rhythm game Arcaea; the author claims it is their first time training a style LoRA and it is only a test version.
• Kurusu Natsume · moss_image_v1.0 — LoRA: A character LoRA for Nijisanji VTuber Kurusu Natsume, MOSS version.
• AuthentiQ Onesie: Jigglypuff [Krea2/MOSS] · MOSS_SFT_v1.0 — LoRA: A realistic onesie clothing LoRA from the AuthentiQ series: makes characters wear a Jigglypuff-themed kigurumi (onesie).
• Sukoya Kana Nijisanji · moss_image_v1.0 — LoRA: The MOSS version of the character LoRA for Nijisanji VTuber "Sukoya Kana."
• Richamu · moss_image_v1.0 — LoRA: A character LoRA for VTuber Richamu, MOSS version: grey hair, small wings on head, red-purple eyes, diamond-shaped pupils.
• GhostFace Mask · v1.0 — LoRA: A prop LoRA for the Ghostface mask from the horror film Scream; the author states it can generate realistic Ghostface masks, available in clean and worn versions.
• Masoq style | Anima | Illustrious · MOSS v1.0 — LoRA: The MOSS version of a black-and-white anime style LoRA named after Twitter artist "Masoq"; this version has no visible examples.
• dorop style moss · v1.0 — LoRA: The MOSS version of artist dorop's personal style LoRA; the model card only states "trained using modelscope," with no textual description of the style.
• Moody Vivid Mix - Moss Image (Free Download) · V2.2 25STEPS CFG3 BF16 — Checkpoint: The BF16 full-precision weights (paid) of the realistic photography integrated checkpoint Moody Vivid Mix V2.2; the author states the series focuses on high-contrast, high-saturation realistic photos and diverse faces. A free quantized version of the same V2.2 is also available.
• (moss-image) 2in1 Kagura Tsuchimiya and Isayama Yomi · v1.0 — LoRA: A dual-character LoRA for Kagura Tsuchimiya and Isayama Yomi from the anime Kurokami -The Second Coming-; one file with two trigger words; the author's only note is "To be improved."
• Kei Mochizuki · Moss v2.0 — LoRA: The second version of the style LoRA named "Kei Mochizuki" for MOSS; the author states it is much better after retraining; the trigger word hint from the Krea 2 version suggests a style with thin lines and flat, low-saturation colors.
• Hoang Trung Kien · v1.0 — LoRA: A LoRA where the model name is a person's name and the description only says "lora dev"; inferred from the filename handcraft-diorama to be a handmade miniature scene style, though the author does not specify.
• Male Flat-Top Hair [MOSS} · v1 — LoRA: A concept LoRA that makes MOSS draw the classic male flat-top / military-style flat top haircut; paid download.
• MOSS R. Dorothy Waynewright (from The Big O) · v1.0 — LoRA: A character LoRA for the android girl R. Dorothy Waynewright from the anime THE BIG O; paid download.

---

Data: Civitai Public API (Base Model MOSS Image), inclusion window 11-01 08:31 → 11-01 21:01 (Beijing Time). Recommendations are editor-reviewed; download counts only reflect attention.