Zhenzhen's AI Radar · MOSS Image · 2026-11-15 Morning Report
This issue features 41 new versions: Strongly Recommended 1 · Recommended 11 · Worth Watching 10 · Included Only 19
Editor's Note This issue reviews 43 new MOSS versions: 1 S, 11 A, 11 B, 20 C, with no exclusions. Top pick: Dizzle's Cloud Photography, a surreal landscape LoRA featuring massive cloud formations dominating the frame, high-contrast black-and-white aesthetics, and clouds tinted in a single color (pink or purple), with full details on trigger words, strength, steps, CFG, and sampler. Other A-tier entries include two other works by the same author (Fish and Lemon), the Tesla Roadster and 2CV concept car, Tiny Things miniatures, the eyes-upward pose, the MOSSquid merged base model, and the browser bounding-box layout tool Layout Studio. Several auto-S entries were downgraded due to missing weights or MOSS-specific parameters. Standard character LoRAs, those with all examples hidden, and those suspected of mislabeling Anima weights as MOSS (Petalmist) were classified as C.
Strongly Recommended (1)
[S] Dizzle's Cloud Photography · v1
LoRA · New Model · Author Maridizzle · Released 11-15 01:25 · Downloads 32
Dizzle's Cloud Photography: A surreal landscape style where the scene is dominated by massive cumulus clouds, rendered in high-contrast black and white, with the clouds tinted in a single color (pink or purple), and ground elements scaled down beneath the clouds.
Images

Reasons for Recommendation
- New model, trained specifically for MOSS by the author; suitable for posters, album covers, atmospheric landscapes, and "small figure against big sky" compositions, with grass, figures, and city skylines placed under the clouds
- Complete parameters: Trigger word, strength 0.9 to 1.0, 30 to 50 steps, CFG 3, and Euler flowmatch are all specified, along with instructions to use full natural language sentences with the trigger word at the beginning
- 5 visible examples (6 hidden); free, with a license permitting commercial use and rental of generated images
How to Use
- Start prompts with
dz_clouds, dense clouds, then describe the scene in full sentences; for the signature effect, addpinkorpurpleto tint the clouds, keeping the rest black and white - Strength 0.9–1.0, 30–50 steps, CFG 3, Euler flowmatch; the author suggests letting clouds occupy most of the frame and scaling down ground scenes
Risks and Limitations
- Inconsistent trigger words: The description uses
dz_clouds, while the site listsdz_cloud; you may need to try both - The black-and-white plus single-color cloud composition is fixed, leading to potential similarity across multiple images; the page contains mixed content
Trigger Word: dz_cloud
Recommended (11)
[A] DC Cyberpunk Style 1 · v1.0 [MossBase]
LoRA · Port · Author dkpc69 · Released 11-15 05:37 · Downloads 32
MOSS version of the DC Cyberpunk Style 1 art style LoRA: Cinematic cyberpunk featuring heavily modified characters, armored streetwear, complex vehicles, massive cities, and industrial atmosphere, with a retro-modern sci-fi film aesthetic.
Images

Reasons for Recommendation
- Port: The same page already has a Krea 2 (Krea2Raw) version; the MOSS version was retrained on MossBase using the same 206 Midjourney images from the author, with Rank 32, 7000 steps, and Ai-Toolkit; the author states it can be used with MossSFT
- Suitable for cyberpunk character designs, vehicles, and city concept art; the author claims all example images include prompts and ComfyUI workflows, suggesting starting with their prompt style before expanding
- 13 visible examples (7 hidden); free, with a license permitting image generation and rental
How to Use
- Include the trigger word
DCCBRPUNKin prompts, describing character modifications, clothing, vehicles, and cities following the author's examples; the author notes that MOSS may require generating 4 images to get one satisfactory result - The author did not provide strength, steps, CFG, or sampler settings for the MOSS version (the suggestion to increase strength to 1.2 applies to the Krea 2 version)
Risks and Limitations
- The author admits in the Krea 2 version description that the dataset included ordinary people and cars, so simple prompts may yield generic results; it is unclear if the MOSS version has the same issue
- The ChatGPT prompt instruction txt file mentioned by the author is not included in the MOSS version files, only the weights; the page contains mixed content
Trigger Word: DCCBRPUNK,
[A] Tiny Things · v1.0 Moss
LoRA · Port · Author Daalis · Released 11-14 22:15 · Downloads 69
MOSS port of Tiny Things: A macro surreal style where subjects are rendered as tiny, fragile miniatures (porcelain dolls, stitched fabric dolls, wire, paper fiber textures) placed in vast, quiet environments, emphasizing material details like cracks, dust, and rust.
Images

Reasons for Recommendation
- Port: The same page already has versions for MiniMax H3, Krea 2, Anima, Flux.2 Klein, Qwen, Flux.1, Z Image Turbo, and Illustrious; the MOSS version uses the same trigger word; the author states training on MOSS was smooth but introduced some MOSS-specific "personality quirks," without detailing the differences
- Suitable for macro/miniature scenes, fairy-tale still lifes, and quiet character vignettes; the author claims it works for animals, people, objects, locations, and full scenes, with simpler compositions and clearer subjects yielding more stable style
- Includes 6 visible examples (no hidden ones); the entire model page has accumulated 4943 downloads and 377 likes, while this version has 69 downloads in 11 hours, indicating only popularity
How to Use
- Include the trigger word
TinyDaalin prompts; the author recommends strength 1.0, with a usable range of 0.5–1.5, where lower strength yields more normal images and higher strength intensifies the style - The author provides additional auxiliary terms:
macro photography, extreme scale contrast, miniature subject, porcelain, stitched fabric, dust particles, shallow depth of field, muted colors, etc., which can be selected as needed; steps, CFG, and sampler settings were not provided by the author
Risks and Limitations
- The MOSS version has only a brief version note, with steps/CFG parameters inherited from older versions but not explicitly provided; the author admits subjects often ("perhaps too often") resemble dolls
- Complex or conflicting prompts weaken the style (according to the author); the page is marked as mixed content
Trigger Word: TinyDaal
[A] Tesla Roadster [Moss] · Moss
LoRA · New Model · Author denrakeiw · Released 11-14 21:11 · Downloads 25
2026 Tesla Roadster concept sports car LoRA: Matte white wedge-shaped body, gold-tinted glass cockpit, butterfly doors, hidden pop-up headlights, four rear cold-air thrusters, and cockpit interior.
Images

Reasons for Recommendation
- New model; suitable for automotive product shots, concept car posters, and sci-fi vehicle scenes; the author claims training on 68 watermark-free photos covering studio shots, night roads/tracks, cockpit, and close-ups
- High state controllability: The author claims it can generate the whole car with doors closed, or with butterfly doors raised and active rear wing deployed; text like license plate
RDSTRand headrestROADSTERshould be written in English double quotes - The model card provides multiple complete example prompts (three-quarter view in dark studio, doors open with wings raised, rear chase shot on night road, rear view under starry sky, cockpit steering wheel, headlight close-up); 13 visible examples (7 hidden), with a license permitting commercial use of generated images
How to Use
- Use
Tesla Roadster Conceptin prompts, starting withCreate a photograph of ...following the author's examples, using full sentences to specify camera angle, lighting, door/wing state, and required text; strength 0.85–1.0 - Steps, CFG, and sampler settings were not provided by the author
Risks and Limitations
- The model card describes it as a Flux LoRA with Flux long-sentence style annotations, but the MOSS version lacks separate documentation
- Trained only on this single car; as a real brand model, commercial use requires attention to trademark issues; the page contains mixed content
Trigger Word: Tesla Roadster Concept
[A] HaDeS MOSS Suite Workflows · HaDeS Surprise Generator
Workflow · Version Update · Author _HaDeS_ · Released 11-15 01:14 · Downloads 37
The newly added "Surprise Generator" ComfyUI workflow in the HaDeS MOSS Suite collection: If left blank, MOSS's built-in text encoder randomly generates a complete prompt for image generation; alternatively, input an idea to have it expanded into a long prompt.
Images

Reasons for Recommendation
- Workflow JSON, not new weights; released as the 7nd version on the same collection page (previous versions included ControlNet, img2img, LoRA toolkit, layout editor, and 8-step upscaling workflows)
- New capability: The seed first selects one of 26 media types (from documentary photography to ukiyo-e, comics, claymation), then selects a subject and era/location, with MOSS's Qwen3.5-9B text encoder writing a scene; no additional language model required
- Suitable for inspiration and batch generation: The author suggests setting batch size to 20 to view results; the same seed reproduces the lottery, prompt, and image, with the generated prompt saved in PNG metadata
How to Use
- Requires HaDeS MOSS Suite v0.7.5 (installable via ComfyUI Manager / Comfy Registry), MOSS-Image text encoder, and flux2 VAE
- Tuned for HaDeS MOSS v1.0 Turbo: 8 steps, CFG 1; the author claims any MOSS-Image model can be used. Leave the idea box blank for random generation, or enter a sentence for expansion
Risks and Limitations
- The author did not write update notes (version notes section is empty); usage instructions are only in the general description on the collection page
- Depends on the author's own node package and Turbo LoRA parameters; 4 visible examples, workflow not verified by running
[A] Uwame -eyes upward pose lora- · MOSS Image
LoRA · Port · Author yakitoritamago · Released 11-15 06:06 · Downloads 37
Eyes-upward (Uwame) pose LoRA: Makes only the pupils look upward while keeping the head and face orientation unchanged, used to create "looking up without raising head" or "eye-roll" expressions.
Images

Reasons for Recommendation
- Port: The same page already has Krea 2 and ZImageBase versions; the author claims the MOSS version was trained on the same dataset
- Solves the issue where writing
looking upraises the entire head, suitable for character expression control: shots requiring separated gaze and head orientation, such as helplessness, eye-rolls, or glancing upward - 7 visible examples, no hidden ones; 37 downloads and 4 likes in 3 hours since release
How to Use
- Prompt includes yt_uwame and eyes upwards; the author states that adding descriptions of rolling eyes or similar terms for looking upward can enhance the effect.
- The author did not provide values for strength, steps, or CFG.
Risks and Limitations
- No recommended strength is provided; you must experiment to determine the intensity of the effect.
- The license does not allow commercial use of generated images or redistribution of derivatives; attribution is required for use.
Trigger Words: yt_uwame, eyes upwards
[A] Dizzle's Fantastic Fish Photography · v1
LoRA · New Model · Author: Maridizzle · Published 11-15 00:51 · Downloads 30
Dizzle's Fantastic Fish Photography: Transforms scenes into soft, green-toned surreal collages—fish floating in the air, butterflies scattering, framed by ornate antique picture frames, evoking a Victorian sense of melancholy.
Images

Reasons for Recommendation
- New model; suitable for surreal illustrations, poetic posters, and gallery-style concept art. The author suggests including one or a combination of signature elements such as fish, butterflies, or antique frames in the scene.
- Parameters provided: strength 0.9–1.0, 30–50 steps, CFG 3, Euler flowmatch; use full natural language sentences with trigger words placed at the beginning.
- 5 visible examples (2 hidden); free, license allows image generation and rental.
How to Use
- Start the prompt with dz_fish (the description lists dz_fish, green surreal painting, while the registered trigger is dz_fish, green-tinted surreal photograph), then describe the scene in full sentences including fish, butterflies, or frames; set strength to 0.9–1.0, steps to 30–50, CFG to 3, and use Euler flowmatch.
- The author recommends maintaining low-saturation greens and soft lighting, avoiding vibrant color schemes.
Risks and Limitations
- There is an inconsistency between the trigger phrase in the description (painting) and the registered trigger word (photograph); the author does not specify which to prioritize.
- The style composition is fixed (green tones with floating fish and butterflies); performance outside these elements is not described; the page contains mixed content.
Trigger Words: dz_fish, green-tinted surreal photograph,
[A] MOSSquid · v0.3
Base Model / Checkpoint · New Model · Author: giko · Published 11-14 22:29 · Downloads 143
MOSSquid v0.3: An anime-oriented base model created by merging four MOSS base models—Hassaku Moss, Moss Luc v3, MiaoMiao Harem Moss v1.1, and Moss SFT—using Block Merge.
Images

Reasons for Recommendation
- New base model; the author lists the merge ratios: Hassaku Moss 40%, Moss Luc v3 20%, MiaoMiao Harem Moss v1.1 15%, and Moss SFT 25%. Suitable for those wanting a ready-made anime-oriented MOSS base model without adjusting merges themselves.
- Parameters provided: negative prompt, sampler ER SDE, scheduler Simple or SGM Uniform, 25 steps or more, CFG 2.5; the author claims compatibility with ComfyUI and Forge Neo.
- The same version includes BF16 full weights, an int8 quantized version, and a VAE file; published 11 hours ago with 143 downloads and 19 likes, making it one of the most notable entries this period.
How to Use
- Replace the MOSS base model loading (use the int8 version if VRAM is tight); the matching VAE is provided on the same page; use sampler ER SDE, scheduler Simple / SGM Uniform, 25 steps or more, and CFG 2.5.
- Negative prompt: worst quality, bad anatomy, watermark, jpeg artifact; no trigger words.
Risks and Limitations
- Only 2 visible examples (1 hidden); the style tendency after merging can only be judged from these two images.
- Version number is v0.3 with no update notes; the license only allows on-site rental, not commercial use of generated images.
[A] Citroën 2 CV [Moss] · v1
LoRA · New Model · Author: denrakeiw · Published 11-14 23:47 · Downloads 25
Citroën 2 CV Concept Car LoRA: A modern version of the 2CV featuring an orange-red body, longitudinal ribbed hood, rollable canvas roof, and yellow checkered cockpit, including car badges, license plate text, and interior details.
Images

Reasons for Recommendation
- New model; suitable for automotive product images, concept car displays, and interior staging scenes. The author states the training set consists of 88 watermark-free studio shots covering various exterior angles, top-down views, open/closed roof states, interiors, and close-ups.
- The author provided 5 complete example prompts (front three-quarters, rear three-quarters, direct rear, convertible interior view, top-down silhouette), with text on the car such as 2CV, La deuche, CITROËN, and Michelin written in English double quotes.
- 5 visible examples (4 hidden); license allows commercial use of generated images and sales.
How to Use
- Write Citroën 2 CV Concept in the prompt, starting with 'Create a photograph of ...' as per the author's examples, specifying camera angle, studio lighting and background, roof open/closed state, and desired text; set strength to 0.85–1.0.
- The author did not provide values for steps, CFG, or sampler.
Risks and Limitations
- The model card title says Flux LoRA, and annotations follow the long-sentence style of Flux; there is no separate description for the MOSS version.
- Only this specific car was learned; both training and examples use studio backgrounds, so performance in complex scenes like street views is not described.
Trigger Words: Citroën 2 CV Concept
[A] Dizzle's Sugar Fruit Photography · v1
LoRA · New Model · Author: Maridizzle · Published 11-15 01:19 · Downloads 32
Dizzle's Sugar Fruit Photography: Dyes the entire image in a single saturated lemon yellow and repeatedly scatters lemons and lemon slices throughout the scene, creating surreal staged photos resembling installation art.
Images

Reasons for Recommendation
- New model, trained specifically for MOSS by the author; suitable for concept posters, fashion/editorial staged shots, and surreal visuals. The author claims the effect is stronger when scenes are more ordinary or bizarre (e.g., dungeons, national parks, people reading under trees).
- Parameters provided: trigger words, strength 0.9 to 1.0, 30 to 50 steps, CFG 3, Euler flowmatch; instructions specify using full natural language sentences with trigger words at the beginning.
- 5 visible examples (3 hidden); free, license allows commercial use of generated images and rental.
How to Use
- Start the prompt with dz_lemon, yellow-cast photograph (the full registered trigger on the site is dz_lemon, yellow-cast surreal photograph), then describe the scene in full sentences, adding lemon slices.
- Set strength to 0.9–1.0, steps to 30–50, CFG to 3, and use Euler flowmatch; the author advises not balancing the yellow color, as full saturation across the image is required for the effect.
Risks and Limitations
- Both the color and lemon motif are fixed, limiting its utility; finding another model is necessary to change colors or fruits.
- The page contains mixed content.
Trigger Words: dz_lemon, yellow-cast surreal photograph,
[A] BB Channel: Fate/Grand Order · MOSS-sft03
LoRA · New Model · Author: RHU_AIstudio · Published 11-15 05:50 · Downloads 10
Fate/Grand Order BB Channel Studio Scene LoRA: Features a pink checkered NEWS BB background screen, a golden news desk with the BB logo, sofas, bulletin boards, and a pink grid floor.
Images

Reasons for Recommendation
- New model; a scene/background LoRA suitable for FGO fan art, BB-hosted program compositions, and character group photo backgrounds, as similar scene assets are rare.
- The author states it was trained on 3D renders of the recreated studio plus anime-style redraws of these renders, totaling 20 images.
- Trigger words and strength are clearly defined; 4 visible examples (3 hidden), all with prompts; license allows commercial use of generated images.
How to Use
- Write bb_channel_(fate) in the prompt, set strength to 1.0; write characters separately or stack character LoRAs.
- If the layout deviates, change the seed and regenerate; the author did not provide values for steps, CFG, or sampler.
Risks and Limitations
- The author notes that positions of objects like the news desk and sofas will shift slightly each time, only approximating the general layout.
- Utility is limited to this specific scene; the version name is MOSS-sft03, and the author does not specify whether it pairs with an SFT or Base model.
Trigger Words: bb_channel_(fate)
[A] The Layout Studio for MOSS-Image - "Draw" your bbox prompt, then paste the JSON in any generator · v1.0.0
Other · New Model · Author: Rexoplanets · Published 11-15 00:02 · Downloads 20
The Layout Studio: A local offline tool for browser-based bounding box layout for MOSS-Image. After drawing bounding boxes for text, objects, and characters, it exports MOSS JSON prompts to paste into any generator running MOSS.
Images

Reasons for Recommendation
- A tool, not new weights; unlike the previous layout editor which only supported Windows, this is a cross-platform pure web tool requiring no ComfyUI or Python. It can be pasted into Forge Neo, Civitai online generation, or any MOSS generator with a prompt box.
- Suitable for preventing character bleed in multi-character scenes, poster text positioning, product images, and storyboard layouts: boxes are automatically converted within MOSS's central coordinate system (yxyx, -500 to 500). It supports reference image alignment, setting style tags and color schemes, importing existing JSON for editing, or exporting plain text prompts for models that do not recognize JSON.
- The author clearly states limitations: positions written in the prompt override box positions; boxes are not pixel-perfect; text boxes are more accurate than object and character boxes.
How to Use
- Download the zip, extract it, and open in a browser (the author also provides GitHub and an online demo); arrange boxes, export JSON, and paste into the generator's prompt box; do not include absolute positions like 'on the left' in individual box descriptions.
- The author recommends CFG no lower than 2.5, otherwise elements may drift; for final generation, follow OpenMOSS recommendations: ≥ 50 steps, 1.5 MP or 2 MP, CFG 3–6.
Risks and Limitations
- Actual placement depends on MOSS's adherence to bounding boxes; the author explicitly states no guarantee of precise positioning according to the boxes.
- Only 2 visible examples; the tool has not been verified in operation.
Worth Watching (10)
[B] Chiaroscuro Moss SFT · BF16 v1.0
Base Model / Checkpoint · New Model · Author: svnx7x_ · Published 11-15 03:13 · Downloads 33
Chiaroscuro Moss SFT: A complete base model for MOSS created by merging two LoRAs (High Landscape, Character Detailer Booster). The author claims it enhances color contrast and adds slight detail, leaning towards anime characters.
Images

[B] V67 Music Prompt · V1 MOSS
LoRA · Ported · Author: angelomaiota · Published 11-14 21:02 · Downloads 27
V67 Music Prompt: A concept LoRA that draws musical notation into the image—featuring staff notation, neumes, mensural notation, solfège, and Curwen hand signs, paired with instrument scenes such as cello, Roman cornu, and synthesizers.
Image

【B】Bite Me: Umbrium Regional Mouths · Moss v1.0
LoRA · Port · Author umbrium2835 · Published 11-15 01:00 · Downloads 32
Regional Mouths: A surreal concept LoRA that adds an extra mouth to specified body parts of a character. Each of the nine body parts has its own trigger word, and you can also specify mouth opening/closing, tongue protrusion, and tooth shape.
Image

【B】SeedraAI SmartphoneSnapshot LoRA MOSS · v1.0
LoRA · New Model · Author fcktisaa · Published 11-15 05:45 · Downloads 71
SeedraAI SmartphoneSnapshot: Enables MOSS to generate casual snapshot-style smartphone photos—daily snapshots, selfies, and natural lighting. The author notes this is an experimental version.
Image

【B】Style: Celscape · MOSS-base v1.0
LoRA · New Model · Author NerissG · Published 11-15 08:28 · Downloads 7
Style: Celscape: An art style LoRA featuring anime characters with a slightly impasto feel paired with photo-realistic backgrounds (the author's original phrase: 3/4 photo background). The author claims it improves output consistency.
Image

【B】Shelby Cobra 427 - 1965 · v2.0 MOSS
LoRA · Port · Author mastermerlin · Published 11-15 07:42 · Downloads 8
A vehicle LoRA for the 1965 Shelby Cobra 427 sports car, MOSS version. Use the trigger word to generate this classic convertible sports car.
Image

【B】Pink ★ Fehe Rin / ぴんく☆ふぇっへ凛 - Art Style · MOSS Image v1.0
LoRA · Port · Author Mt_Damn · Published 11-15 00:42 · Downloads 74
An art style LoRA based on pixiv artist ぴんく☆ふぇっへ凛 (Pink ★ Fehe Rin), MOSS version. The author states they have fully migrated to MOSS and provides a complete recipe.
Image

【B】Vintage AFO leg brace · v1.0
LoRA · New Model · Author droidy001142 · Published 11-15 02:12 · Downloads 22
A vintage ankle-foot orthosis (AFO leg brace) object LoRA: Adds one or two legs to a character with old-fashioned orthotic braces featuring straps and buckles.
Image

【B】PixelArt [Minimax / MOSS / Krea2 / SD] · MOSS v1.0
LoRA · Port · Author TeraWatt · Published 11-15 06:17 · Downloads 1 · Paid
MOSS version of PixelArt: A 8-bit / 16-bit retro game pixel art style LoRA, used for creating characters, enemies, sprites, scene maps, icons, and items.
Image

【B】Multi LoRA Regional Node · v1.1
Other · New Model · Author Generation3dX · Published 11-14 23:41 · Downloads 0 · Paid
Multi LoRA Regional Node: A ComfyUI custom node that allows multiple character LoRAs to apply only within their respective bounding boxes, enabling multiple distinct characters in a single image.
Image

Included Only (19)
- Flonne (Disgaea) · MOSS — LoRA: Character LoRA for Flonne, the angel from the original Disgaea: blonde hair with a headband, blue eyes, angel wings, white shirt with a red skirt, and white Mary Jane shoes.

- [Moss/Krea2]LzzyHale style · moss v1.1 — LoRA: Second version of the MOSS LoRA named "LzzyHale style" (moss v1.1); the model card does not specify whether it learns character appearance or art style.

- Nobuteru Yuki style · MOSS — LoRA: Art style LoRA named after anime character designer Nobuteru Yuki, MOSS version; the model card contains no textual description.

- Angelina NordhavenLove · MossImage — LoRA: Character LoRA for the fictional character Angelina NordhavenLove (part of the author's Nordhaven family series), set as an art curator on the west coast of Sweden.

- waxa Cinematic Style · v1 — LoRA: waxa Cinematic Style: a cinematic style LoRA; the model card only includes a one-sentence introduction and recommended strength.

- Ada Wong Tactical Outfit (RE4 Remake) | Costume LoRA · v1.0 MOSS — LoRA: Costume LoRA for Ada Wong's tactical outfit from Resident Evil 4 Remake: a red ribbed high-neck knit dress with black leather tactical straps and a holster, applicable to any character.

- Ariel - The Little Mermaid (1989) - Disney - MOSS LORA · v1.0 — LoRA: Character LoRA for Ariel from Disney's The Little Mermaid (1989), trained on MOSS, including descriptions for both human and mermaid outfits.

- Zelda from The Legend of Zelda : Skyward Sword [Krea 2/MOSS Image] · MOSS — LoRA: Character LoRA for Zelda from The Legend of Zelda: Skyward Sword: blonde hair, straight bangs with a hime cut, blue eyes, and pointed elf ears.

- Pocahontas - Disney - MOSS LORA · v1.0 — LoRA: Character LoRA for Pocahontas from Disney's Pocahontas, trained on MOSS, including descriptions for the character herself and her fringed one-shoulder dress.
- Petalmist · v1.0 - Anima — LoRA: Second version of the Petalmist (Flower Mist Soft Painting) art style LoRA, but both the version name and filename say Anima, suggesting Anima weights were mistakenly labeled as MOSS.

- Princess Zelda from The legend of Zelda : Ocarina of time Remake [Krea 2/MOSS Image] · MOSS — LoRA: Character LoRA for Princess Zelda from The Legend of Zelda: Ocarina of Time Remake (MOSS version), the author's first locally trained model.

- azuchi_(azur_lane)/azuchi/安土/あづち/あづち(アズールレーン) · Moss-heal v1.0 — LoRA: Character LoRA for Azuchi from Azur Lane: long purple hair, black curled horns, and a mole below the mouth, including two outfits: the original patterned kimono and a red backless evening gown skin.

- Kigurumi Onesie: SpiderGwen [Krea2/MOSS] · MOSS_v1.0 — LoRA: MOSS version of the Spider-Gwen themed Kigurumi onesie costume LoRA, part of the author's onesie series.

- FymriE style 【krea2&MOSS&anima&il】 · Moss-heal v1.0 — LoRA: New MOSS version of the FymriE art style LoRA, this time trained on the community fine-tuned base model MOSS-heal, with no trigger word.

- Jealous M(o)ess · v1.5_ep 20 — LoRA: Second MOSS version of Jealous M(o)ess: the "evil twin" version of the author's jealous series art style on Tensor, tagged as male-oriented and niji style.

- Kigurumi onesie: Hulk [Krea2/MOSS] · MOSS_SFT_v1.0 — LoRA: MOSS version of the Hulk themed Kigurumi onesie costume LoRA, part of the author's onesie series.

- Natasha_Akt · Natasha_Akt-MOSS — LoRA: Character LoRA for the fictional character Natasha_Akt, which the author claims is entirely created from generated images; there was previously a Krea 2 version, now supplemented with a MOSS version.

- denia in moss · v1.0 — LoRA: MOSS character LoRA for the character "Denia," which the author claims was trained based on MOSS Base.

- Realistic Pumped Veins-Krea2 / MOSS / Anima / ZIB / Ernie-GMR · RealisticPuped veins-MOSS — LoRA: MOSS version (based on MOSS-SFT) of a realistic effect LoRA that makes veins under the skin prominently bulge.

Data: Civitai public API (base model MOSS Image), inclusion window 11-14 20:31 → 11-15 09:01 (Beijing Time). Recommendation reasons have been reviewed by editors; download counts only reflect attention.