MOSS Image Morning Report 2026-11-05 | Top Pick: cocoamixMOSS (moss image)

MOSS Image 2026-11-05 Morning Report: 40 new versions, strongly recommended 7, recommended 19, worth watching 12, included only 2.

2026-11-05T01:53:45Z

Zhenzhen's AI Radar · MOSS Image · 2026-11-05 Morning Report

This issue features 37 new versions: Strongly Recommended 1 · Recommended 12 · Worth Watching 14 · Included Only 10

Editor's Note This issue reviews 42 new MOSS versions. Top pick: Natural Faces v2: a facial detail slider that works in both positive and negative directions; positive values add pores, freckles, and fine lines, while negative values smooth the skin. Trigger words and positive/negative intensity ranges are provided. Tier A 12 items: base models cocoamixMOSS BF16 Turbo and MOSS_LUC V3 FP8, art styles including Disney 2D, Jima, tokkyu, and Saturn, as well as geisha attire, Lada Niva vehicles, and Aetherwild fantasy scenes.

Strongly Recommended (1)

[S] Natural Faces · v2

LoRA · New Model · Author fyrtyv · Published 11-05 06:30 · Downloads 47

A facial detail slider LoRA (MOSS Image SFT) that works in both positive and negative directions: positive values add pores, freckles, and fine lines, as if removing a beauty filter; negative values make the skin smoother and softer.

The use case is clear in one sentence. Trigger words, starting intensity, positive/negative limits, and testing ranges are all provided, making it suitable for same-prompt comparisons at -0.65 / 0 / +0.5 / +1. The main drawback is the limited number of examples.

Reasons for Recommendation

New model: Solves the issue of overly heavy or light "filter effects" in MOSS portraits. The author emphasizes that it changes details during generation, rather than applying post-processing sharpening or skin smoothing.
Serves realistic portrait text-to-image generation: With the same prompt, simply adjusting the intensity allows switching between "realistic skin texture" and "soft-focus smoothness" without changing the base model.
Actionable: Trigger word naturskinx, base model MOSS Image SFT, starting intensity +0.5, recommended positive upper limit 0.75-1.0, recommended negative limit down to -0.65, author's tested range from −1 to +1.
Version name v2, the author states the entire model was remade; free download.

How to Use

Base model MOSS Image SFT; add trigger word naturskinx to the prompt.
Start intensity at +0.5; increase towards 0.75-1.0 for more skin detail, or adjust to negative values for smoother skin, recommended not lower than -0.65 (the author notes that effects converge around negative values near -1.0).
Steps and CFG were not provided by the author; use standard SFT settings.

Risks and Limitations

Only 3 visible examples are available; the comparative effects of positive and negative intensities rely mainly on the author's description.
The license does not allow selling generated images (only renting the model is permitted); only v2 remains on the page, v1 is no longer visible, making it impossible to compare changes; this item was not tested.

Trigger Words: naturskinx

Recommended (12)

[A] cocoamixMOSS (moss image) · BF16_turbo_v1.0

Base Model / Checkpoint · Version Update · Author ccooccoo · Published 11-05 03:21 · Downloads 29

A fused base model for the cocoamix series on MOSS. This version is the BF16 Turbo edition with the 8-step DMD2 acceleration LoRA directly fused in, generating images in approximately 10 steps; the default cocoamix art style does not draw pupils.

Reasons for Recommendation

Version update: The same model previously had a standard BF16 version (50 steps, CFG 4) and an older INT8 10-step version; according to the author's fusion list, this version is MOSS-Image + author's self-trained model + 8-steps-dmd2-lora, belonging to the "acceleration LoRA-free" fast version.
Serves anime-oriented text-to-image generation, with tags also including realistic, mecha, and chibi styles; the author provides two style switches: for pupils, write "eyes has dark_circle shaped pupils"; for vivid colors, write "vivid dark color"; realistic images have specific example prompt sentence structures.
Parameters are complete: 10 steps, CFG 1 - 4 (CFG 1.0 is fastest, negative prompts can only be used above 1.5), samplers ER SDE + simple / Euler + normal / DDIM + normal; 10 example images.

How to Use

Switch the base model to cocoamixmossMossImage_bf16TurboV10.safetensors (BF16 full weights), no trigger word.
10 steps; CFG 1.0 for speed, adjust to above 1.5 if using negative prompts; sampler ER SDE + simple or Euler + normal.
Add "eyes has dark_circle shaped pupils" for pupils; add "vivid dark color" for high saturation; use the sentence structure "photorealistic style has realistic background, realistic hair, realistic face..." for realistic photos.

Risks and Limitations

The author did not write update notes (the version description field is empty); the difference between this version and the standard version can only be inferred from the version name and the model card's fusion list.
Strict license: Commercial use is prohibited, and generated images are also limited to non-commercial use; fused releases and selling generated images require prior consent from the author, and reposting to unauthorized sites incurs fees.
The author warns that sensitive content is easily generated, suggesting the use of prompt_enhance for control; this item was not tested.

[A] Saturn - Artist Style · MOSS-Image

LoRA · Ported · Author Azriaelis · Published 11-05 00:05 · Downloads 46

A semi-realistic illustration style LoRA imitating artist Saturn (@saturn_(satscarlet)), following the high-saturation, detail-dense path of League of Legends splash art, ported from Krea 2 / Anima to MOSS.

Reasons for Recommendation

Ported: The same art style previously had Anima and Krea2Turbo versions; this adds the MOSS-Image version and provides separate recommended parameters for MOSS (one set for each of the three base models).
Serves game splash art-style character illustrations and semi-realistic digital painting text-to-image generation.
The author states that the style is significantly stronger when the trigger word is included; MOSS parameters: Pair with Base + 8-step Turbo LoRA, euler + simple, LoRA strength 0.8-1.2, 8-16 steps, CFG 1-3; 12 visible examples.

How to Use

Write trigger words splash art, league of legends, vibrant colors, @saturn_(satscarlet) in the prompt.
Use MOSS Base as the base model and attach the 8-step Turbo LoRA; euler + simple; LoRA strength 0.8-1.2; 8-16 steps; CFG 1-3.

Risks and Limitations

Imitates the personal art style of a specific living artist and uses the artist's ID and game name directly as trigger words; commercial use of generated images requires self-assessment.
8 sensitive-level examples are hidden; the license does not allow releasing derivative/fused models; this item was not tested.

Trigger Words: splash art, league of legends, vibrant colors, @saturn_(satscarlet)

[A] Disney Style - MOSS · V1

LoRA · New Model · Author vizsumit · Published 11-05 02:52 · Downloads 156

A LoRA that renders images in the classic Disney 2D hand-drawn animation style: fluid lines, expressive characters, and soft hand-drawn backgrounds.

Reasons for Recommendation

New model: Disney-style traditional cel animation art style on MOSS; the author states it is suitable for characters, portraits, fantasy scenes, and fairy-tale-like environments.
Simple usage: A fixed opening trigger word phrase + intensity 0.6 – 0.8; the author says the workflow and prompts are in the example images, with 15 visible examples.
Published about 7 hours ago with 156 downloads, making it one of the more downloaded items (indicating attention level only).

How to Use

Start the prompt with "This Disney-style digital illustration depicts" followed by the scene description.
LoRA intensity 0.6 – 0.8; steps, CFG, and whether to use Base or SFT were not written in the description by the author, requiring reference to the workflow attached in the example images.

Risks and Limitations

Directly named after Disney and imitating its art style; although the license states that selling is allowed, commercial use of generated images requires self-assessment of copyright and trademark risks.
The description is very short; inference parameters can only be found in the example image metadata; 5 sensitive-level examples are hidden; this item was not tested.

Trigger Words: This Disney-style digital illustration depicts

[A] Blackwall : Brutalist Dystopian Cyberpunk · MOSS

LoRA · Ported · Author NeuralVandal · Published 11-05 08:19 · Downloads 18

A dark cyberpunk atmosphere LoRA based on the "Blackwall" concept from Cyberpunk 2077: brutalist concrete megastructures, industrial machinery, neon, and digital decay, ported from Krea 2 to MOSS.

Reasons for Recommendation

Ported: Previously had a Krea 2 version; the MOSS version, based on the filename, was trained on MOSS Base.
Serves oppressive, heavy dystopian sci-fi scenes and concept design text-to-image generation; the author states it is more of a visual concept than a complete art style and can be stacked with other style LoRAs.
Trigger words are fully registered (Cyberpunk style, Blackwall, neon-infused, digitally crafted, futuristic), with 13 visible examples.

How to Use

Write Blackwall, Cyberpunk style in the prompt, optionally adding neon-infused, digitally crafted, futuristic.
LoRA strength, steps, and CFG were not provided by the author.

Risks and Limitations

Inference parameters were not provided by the author, and the description largely discusses atmospheric concepts; 7 sensitive-level examples are hidden.
Based on the setting from the game Cyberpunk 2077; commercial use of generated images requires self-assessment; this item was not tested.

Trigger Words: Cyberpunk style, Blackwall, neon-infused, digitally crafted, futuristic

[A] Geisha / 芸者 · Moss-Image

LoRA · New Model · Author Takec5031 · Published 11-05 00:59 · Downloads 46

A LoRA for drawing "true geisha" (geisha / geigi) attire: A Japanese author specifically distinguishes geisha from maiko, oiran, and princesses, which are often mixed up by AI.

Reasons for Recommendation

New model: The author states that when AI draws ancient Japanese women, it often mixes up the hairstyles and attire of geisha, maiko, oiran, and princesses. This LoRA focuses only on geisha—mature hairstyles and ornaments, white makeup with red lips, furisode kimono, and obi.
Serves Japanese-style character, historical theme illustrations, and realistic images; the author clearly explains the visual differences between the four identities in the description, which can serve as prompt references.
Trigger words and a full set of attire tags are provided; 1 example video segments + 17 example images.

How to Use

Write geisha in the prompt, followed by the tag string provided by the author: traditional hair, nihonkami, hair ornament, geisha makeup, classic geisha updo, white foundation, red lips, kimono, furisode, long sleeves, obi, obijime, susohiki, white tabi, holding folding fan, sensu.
LoRA strength, steps, and CFG were not provided by the author.

Risks and Limitations

Inference parameters were not provided by the author, and the applicability is limited to this specific traditional attire; this item was not tested.

Trigger Words: geisha, traditional hair, nihonkami, hair ornament, geisha makeup, classic geisha updo, white foundation, red lips, kimono, furisode, long sleeves,obi,obijime, susohiki, white tabi, holding folding fan, sensu,

[A] Niva Legend (VAZ-2121) · v1.0 MOSS

LoRA · New Model · Author bazya022020810 · Published 11-05 05:41 · Downloads 15

A LoRA for drawing Soviet off-road vehicles, the Lada Niva (VAZ-2121). The author states that the training set includes both traditional off-road lifestyle scenes and cyberpunk modified vehicle scenes.

Reasons to Recommend

New model: A single-vehicle LoRA on MOSS. The author emphasizes accurate body proportions and seamless integration with the environment.
Serves realistic automotive photography, off-road scenes, and cyberpunk/post-apocalyptic modified vehicle concept art.
Trigger word: niva legend (mentioned in the description but not registered in the trigger word field). Strength range 0.7 – 1.0. Includes two complete example prompts (dusk by a forest river, neon-lit night city modified car); 9 visible examples.

How to Use

Start the prompt with niva legend, then add environmental descriptions such as rough gravel, forest river, or neon-lit cyberpunk megacity.
LoRA strength: 0.7 – 1.0; steps and CFG were not provided by the author.

Risks and Limitations

The author labels this as a beta version and warns that MOSS may not generate images as expected even when the LoRA is applied.
The license does not permit selling generated images or publishing derivative models; this entry has not been tested.

【A】Jima Style Anima&Krea 2&Moss · Moss

LoRA · Port · Author: storyAura · Published 11-05 01:19 · Downloads 49

A MOSS version of the Jima-style anime illustration LoRA, ported from Krea 2 / Anima. The same page offers three files trained with different step counts for selection.

Reasons to Recommend

Port: Previous versions existed for Anima (Anima Base-v1) and Krea 2 (Krea 2 Raw). The MOSS version is trained based on Moss-Image-Base-bf16.
Serves text-to-image generation of anime girl illustrations. The three files on the same page have different training levels; the author recommends the one with 3500 steps (jima-comfy-3500.safetensors), but users can also compare them themselves.
The author suggests using Base + dmd2 distillation for accelerated inference, claiming it yields better results than using SFT. Two parameter sets are provided: without dmd, 30-50 steps, CFG 3-5; with dmd, 8 steps, CFG 1. Sampler: Euler + SGM or Euler + Simple; resolution from 512^2 to 1536^2.

How to Use

Preferred file: jima-comfy-3500.safetensors. No trigger word; describe the scene by referring to the example prompts.
Base model: MOSS Base. When using dmd for acceleration, use 8 steps and CFG 1; without it, use 30-50 steps and CFG 3-5. Sampler: Euler + SGM / Simple.
LoRA strength was not provided by the author.

Risks and Limitations

Only 4 visible examples are available. The author notes that the generation parameters are "quoted from official documentation" and were not tuned specifically for this LoRA.
Mimics a specific artist's style. The author states that sensitive-level content is partially suppressed; this entry has not been tested.

【A】MOSS_LUC [ moss-image-checkpoint ] · sft-fp8-v3

Base Model / Checkpoint · Version Update · Author: ukr8b3g201 · Published 11-04 22:19 · Downloads 19

The V3 FP8 version of MOSS_LUC, an anime-oriented MOSS fused base model. The author claims it stabilizes the general scheduler, adjusts contrast, and makes the style "cuter," focusing on anime illustration rather than realism.

Reasons to Recommend

Version Update: Changes in V3 are detailed in the model card—stabilizing the general scheduler, adjusting contrast, and leaning more towards KAWAII. This version is the FP8 file of V3. The V3 INT8 version (sft-int8-v3) was released the previous day, and before that was the V2 INT8 version.
Serves text-to-image generation in anime and illustration styles. The author explicitly states that realism is not a strength.
Full parameters provided: DPM++ SDE or DPM++ 2M SDE, Beta or MOSS Scheduler (Shift 5). Steps from 8 to 10, starting at 8. CFG from 2 to 5, starting at 3. Paired with the DMD acceleration LoRA lora_latest.safetensors at strength 0.7–0.8. Recommended positive and negative prompts are included, along with 12 visible examples.

How to Use

Switch the base model to mossLUCMossImage_sftFp8V3.safetensors (FP8, free)
Apply the DMD acceleration LoRA lora_latest.safetensors (lodestones/MOSS-Image-SFT-8-steps-dmd) with a strength of 0.7–0.8; the author uses 0.8
Use 8–10 steps (starting from 8) and CFG 2–5 (starting from 3, lower for photorealistic styles); install ComfyUI-MOSS-Shift-Scheduler in ComfyUI or the corresponding MOSS Scheduler extension in Forge Neo, select Shift, and set the value to 5
Start positive prompts with masterpiece, best quality, amazing quality; use negative prompts such as worst quality, low quality, lowres, bad anatomy, bad hands, etc.

Risks and Limitations

The author did not write update notes (the version description field is empty). FP8 and the previous day's INT8 are different quantization formats of the same V3, not new capabilities.
The author recommends using their own scheduler extension, which relies on a third-party node. The author states that realism is not a strength; if forcing realism, keep CFG between 2 and 3.
2 sensitive-level examples are hidden; this entry has not been tested.

【A】Aetherwild Fantasy Environments · Aetherwild MOSS 800 steps

LoRA · New Model · Author: CelestialCanvas · Published 11-05 04:42 · Downloads 30

A fantasy environment LoRA: Use one master trigger word to define the overall style, then use nine zone-specific trigger words to switch between locations such as magical forests, elven courts, ruined castles, crystal caverns, and dwarven forges. The images do not contain characters, leaving space for them.

Reasons to Recommend

New Model: Specializes in environments rather than characters. The training set deliberately excludes people and creatures. The author designed it to be stacked with character LoRAs (generate the scene first, then apply the character LoRA and its trigger word).
Serves text-to-image generation for fantasy illustrations, game concept scenes, and realistic fantasy backgrounds. Each of the nine locations has an independent trigger word, offering better directional control than a single style word.
Training information is public: 72 environment images, nine location categories, 800 steps, rank 16, alpha 16, BF16 (AI Toolkit).

How to Use

Write aetherwild in the prompt to define the overall style, then add one zone-specific word, such as aetherwild_enchanted_forest, aetherwild_crystal_caverns, aetherwild_dwarven_forge_halls, or aetherwild_moonlit_sanctuary.
To include characters, apply a separate character LoRA and its trigger word. The author did not provide LoRA strength, stating it should be adjusted based on the base model and prompt; steps and CFG were also not provided.

Risks and Limitations

Only 3 visible examples are available; it is impossible to have examples for all nine zones. It is not specified whether Base or SFT was used.
Fewer training steps and small data volume mean generalization capability is unknown; this entry has not been tested.

Trigger Words: aetherwild, aetherwild_enchanted_forest, aetherwild_fae_blossom_gardens, aetherwild_elven_palace_courts, aetherwild_ancient_ruined_citadel, aetherwild_arcane_observatory, aetherwild_crystal_caverns, aetherwild_dwarven_forge_halls

【A】xSP · MOSS Image

LoRA · New Model · Author: randomimagepotato · Published 11-05 01:08 · Downloads 46

A dark steampunk style LoRA: Features brass, gears, and clockwork elements with dim, soft lighting. All examples are indoor scenes (living rooms, bathrooms, airship cabins, bars, stages).

Maintained at A by automatic A: The style is instantly recognizable, examples are sufficient, and prompts can be copied directly; the same author's Solarpunk was rated B due to fewer visible examples, while this one has more.

Reasons to Recommend

New Model: The author describes it as a high-detail, dark soft-light steampunk style with clockwork and steam elements. All training images are AI-generated; no real photos were used.
Serves text-to-image generation for concept design, indoor scenes, and atmosphere shots. 6 visible examples, all with prompts attached; no hidden examples.
Two trigger words are registered: the short trigger xSP and a full-sentence atmosphere description. The author states that examples were intentionally generated using different MOSS fine-tuned base models to demonstrate differences between them.
Same author and writing style as the Solarpunk LoRA included in the previous issue; free. Published approximately 9 hours ago with 46 downloads (indicating attention only).

How to Use

Load xsp_MOSS_v1.safetensors. Start the prompt with "xSP," and optionally add the full sentence: "The overall ambiance is warm, cozy, and slightly secret, blending industrial and vintage elements with a touch of magic."
The author's commonly used base models are Jib Mix MOSS Realistic and HyperMOSS Lush Turbo. Strength, steps, CFG, and sampler were not provided by the author.

Risks and Limitations

The author warns that MOSS is not very stable even with LoRA applied, leading to variations in output.
Example image metadata (not author instructions): 9 steps, CFG 1, Euler + beta. Base models are all third-party fine-tunes (Jib Mix MOSS Realistic, HyperMOSS Lush Turbo); performance on official Base / SFT is unknown.
All examples are indoor scenes without people; there are no examples for characters or outdoor scenes. The license does not permit publishing derivative models; this entry has not been tested.

Trigger Words: xSP, The overall ambiance is warm, cozy, and slightly secret, blending industrial and vintage elements with a touch of magic

【A】MOSS Anime Classic HD · v2.0

LoRA · Version Update · Author: _Envy_ · Published 11-04 22:28 · Downloads 86

Version v2.0 of the classic cel-shaded anime style LoRA. The author states it was retrained with more refined, color-adjusted training images, resulting in sharper details and more saturated colors than the previous version.

Maintained at A: The direction of changes is clear, examples are sufficient to view, and samplers/accelerated base models are provided. However, strength is not specified, and there are many hidden examples.

Reasons to Recommend

Version Update: The version description field is empty, but the model card adds a note about v2.0 changes—retrained with more detailed images and unified color grading for a more vibrant look, making it clearer and more saturated than the original.
Serves text-to-image generation of anime character illustrations. Tags include flat coloring, cel shading, and high detail. The author recommends pairing with KROMOSS turbo for output.
9 visible examples, all with prompts attached. Prompts uniformly start with a style description sentence that can be copied directly.
Free; the license permits image generation and derivatives. This version has 86 downloads, with cumulative model downloads at 523 (indicating attention only).

How to Use

Load moss-classic-anime-hd.safetensors. Pair the base model with KROMOSS turbo. Sampler: res_multistep + simple (author recommended).
v2.0 has no registered trigger words (the previous version registered anime style / cel shaded). Example prompts all start with "vibrant cel-shaded anime with crisp, fine details and rich, saturated colors," followed by tags.
Strength was not provided by the author; steps and CFG are only found in example metadata (see Risks).

Risks and Limitations

The author did not write update notes (the version description field is empty); changes are only mentioned in a single sentence on the model card.
11 sensitive-level examples are hidden, exceeding the number of visible examples. Visible examples are mostly female character portraits.
Example image metadata (not author instructions): 8 steps, CFG 1.0, res_multistep + simple. Base model name was not recorded; KROMOSS turbo requires separate download, and the author did not provide a link; this entry has not been tested.

【A】Art Style：tokkyu · v1.0

LoRA · New Model · Author: Suzuka_naoko · Published 11-04 23:37 · Downloads 45

Pink-purple soft-light anime-style LoRA (named Rosy Haze by the author): The author describes an overall layer of pale pink light haze, with fine lines, light hair strands, and prominent blush, suitable for quiet solo subjects such as school settings, winter scenes, and night views.

The description format follows the same template as the previous issue's sy4 Clear Sky; training configurations, two sets of sampling parameters, and prompt templates are all provided, so you can generate images by copying them directly; what is missing is that there is only one visible example.

Reasons for Recommendation

New model: The author states it was trained on 36 works from 2023 onwards using MOSS-Image Base (excluding line art and posters with large text), annotated with WD14 English tags, rank 32 / alpha 128, and the released version is the author's recommended e112 checkpoint (4032 steps)
Serves text-to-image generation for anime solo illustrations; the author explains the base model differences: SFT has a heavy built-in style that pulls the effect back, so use Base for a more pronounced style, or add DMD on top of SFT (for lighter and brighter colors)
Parameters are complete and divided into two sets: one without DMD and one with a 8-step DMD acceleration LoRA, each providing steps, CFG, sampler / scheduler, along with strength, resolution, negative prompts, and three complete example prompts
Free; released about 10 hours ago with 45 downloads (indicating attention only)

How to Use

Load tokkyu_e112.safetensors, strength 1.0; write "@tokkyu" at the beginning of the prompt, using the English tag format "@tokkyu, 1girl, solo, ..."; the author states it works without the trigger word, but including it is more stable
Without DMD: 40-50 steps, CFG 3, euler / simple, shift 5
With 8-step DMD acceleration LoRA: 12 steps, CFG 1, euler / simple
Resolution approx. 1MP, e.g., 1024×1536, 832×1216; negative prompt: worst quality, bad anatomy, watermark, jpeg artifact, blurry; if you want text to appear on fans or signs, include the text content in the prompt

Risks and Limitations

Only 1 visible examples (another 2 sensitive-level examples are hidden), and the examples do not include generation parameters, so the style relies mainly on the author's description
The trigger word is not registered on the site, only mentioned in the description; the author states all training images are solo shots, so multi-person scenarios are unknown
tokkyu is an artist's name; the author declares it is unofficial and asks not to impersonate the artist or use it commercially, but the site license checkbox allows selling, creating an inconsistency; handle commercial use according to the author's declaration; this entry was not tested

Worth Watching (14)

[B] [MOSS-Image] KomoreBeams · beta1.0

LoRA · New Model · Author 9d6_Esc · Released 11-05 06:49 · Downloads 30

Judging by the name, this is a beta version of a style LoRA (KomoreBeams) featuring "komorebi"-style light beams filtering through trees; the author directly retrained on MOSS using reference images and annotations from the Krea 2 version, but the description does not detail the effect itself.

[B] RLY's MOSS Realism Workflow · v1.0

Workflow · New Model · Author RaymondLuxuryYacht · Released 11-04 21:37 · Downloads 105

An introductory ComfyUI workflow (JSON) for MOSS realistic portraits; the author states it is the first version, providing two sets of parameters for high and low VRAM; it is not a new weight.

[B] Lithe (Waist Slider for MOSS, Krea 2, Z-image) · MOSS v1.0

LoRA · Port · Author SloopDeeDoop · Released 11-05 00:02 · Downloads 91

The MOSS version of the waist slider LoRA: positive values make the waist thinner than the "slim waist" prompt, while negative values make the waist and abdomen rounder; the author states other body parts are largely unaffected.

[B] Qipao (MOSS) · v1.0

LoRA · New Model · Author EDG · Released 11-05 04:50 · Downloads 27

Qipao clothing LoRA: realistic Shanghai-style qipao looks; examples feature qipao-wearing women with short hair, updos, and pearl earrings.

[B] xCorruptedCore · MOSS Image

LoRA · New Model · Author randomimagepotato · Released 11-04 22:42 · Downloads 39

The author's original "CorruptedCore" style LoRA: cottagecore interiors and villages are corrupted—featuring eerie furniture with organic forms, distorted windows, and elements of sulfur and hellfire.

[B] Kigurumi Onesie: Frog [Krea2/MOSS] · MOSS_SFT_v1.0

LoRA · Port · Author AuthentiQ · Released 11-05 00:06 · Downloads 36

The MOSS SFT version of the realistic frog kigurumi onesie clothing LoRA, which can be stacked on character LoRAs to dress characters in this outfit.

[B] Yang Shu Style [Anima/Krea2/MOSS] · MOSS v1.0

LoRA · Port · Author Imagine__That · Released 11-04 21:17 · Downloads 67

The MOSS version of a LoRA imitating illustrator Yang Shu's style; the author states it generates semi-realistic handsome men and beautiful women.

[B] HELIOS-7 Xenobiology Facility Environments · MOSS-Image-SFT — 800 steps

LoRA · New Model · Author CelestialCanvas · Released 11-04 22:30 · Downloads 63

Dark realistic environment LoRA: abandoned underground xenobiology experiment / containment facilities—brutalist concrete, rusted steel, giant cylindrical bioreactors, teal specimen lights, and red emergency lights; generates scenes only, no people.

[B] Moss/Anima-Style-五更诗音 Gokou shionn · Moss-base-V1.0

LoRA · Port · Author hardworkingsaltedfish · Released 11-05 00:39 · Downloads 70

The MOSS version of a character illustration style LoRA by artist Gokou shionn, known for glossy character art: the author states it was commissioned training, released with the artist's consent, characterized by clean, soft line art and glossy skin with highlights.

[B] Porsche 911 Turbo Targa (930) - 1989 · v1.0 MOSS

LoRA · Port · Author mastermerlin · Released 11-05 06:00 · Downloads 12

The MOSS version of a car model LoRA for the 1989 Porsche 911 Turbo Targa (930), used to generate realistic photos of this car.

[B] Moss_Wheelchair · v1.0

LoRA · New Model · Author droidy001142 · Released 11-04 23:30 · Downloads 56

A modern manual wheelchair concept LoRA: places characters sitting in or pushing wheelchairs in everyday scenes like theater lobbies and streets.

[B] MOSS 12 Step Turbo · v1.0

LoRA · New Model · Author anyMODE · Released 11-05 04:47 · Downloads 65

An acceleration LoRA for MOSS Image: the author describes it as a "compromise" Turbo, recommending 12 steps for generation, though fewer steps may also work.

[B] Sam_MOSS · Turbo v2.7

Base Model / Checkpoint · Version Update · Author toya_san · Released 11-05 04:57 · Downloads 125

Sam_MOSS base model's Turbo v2.7: an "accelerated" full model based on the INT8 version; the author states it is usable at 8 steps and above, and there will be no BF16 version.

[B] Yahata_Umiri_MOSS_lora · v1.1a

LoRA · New Model · Author yueyinganshuang · Released 11-05 01:35 · Downloads 21

A character LoRA for Yahata Umiri: provides complete prompts for four outfits—casual wear, summer uniform, autumn uniform, and stage costume—switching clothes via outfit keywords.

Included Only (10)

Suzan : Classic Hollywood Bombshell · MOSS — LoRA: A character LoRA named Suzan (tags include ai model, likely a virtual model), following the classic Hollywood bombshell (pin-up) route, ported from Krea 2 / Anima / Illustrious / Flux.1 D to MOSS.
Gemma (Monster Hunter Wilds) | Krea 2 + MOSS · MOSS — LoRA: The MOSS version of a character LoRA for Gemma from Monster Hunter Wilds; the author states it is primarily realistic / semi-realistic but can also be prompted for anime style.
Gokurakugai | Yuto Sano style · MOSS v1.0 — LoRA: The MOSS version of a LoRA imitating the style of Yuto Sano, author of the manga Gokurakugai; the model card contains only the title line.
moon style (for sft) · v1.0 — LoRA: An anime style LoRA named moon_style; the author states it was trained on MOSS SFT BF16 and performs better on SFT than Base, uploading both 1250-step and 1500-step files.
Remsrar style · MOSS v1.0 — LoRA: The MOSS version of the style LoRA "Remsrar style"; tags indicate a brushstroke-feel anime style, and the model card text consists only of the name line.
Nagabe/Mucknagabe style · v0.1 MOSS-Image — LoCon: A LoKr imitating the style of artist nagabe (mucknagabe); the author states training images were scattered illustrations without long-form manga content, and feels the result is "a bit off".
Hip dips & thick thighs · v1.0 — LoRA: Body type LoRA: enhances the character's "hip dips" and thick thigh physique; the author states it was trained on an anime dataset.
hutao moss image lora · hutaov moss1.0 — LoRA: A character LoRA for Hu Tao from Genshin Impact; this is the author's first time training a character LoRA, and they admit the style is unstable with significant variation between generations.
Popogori Style · MOSS — LoRA: The MOSS version of a LyCORIS for artist Popogori's 3DCG style; previously only a Pony version existed.
Moss Mafia · Westside — Base Model / Checkpoint: The second version "Westside" of the MOSS base model Moss Mafia; the author states outputs resemble Flux 2 Klein somewhat, but hands and legs are not good.

Data: Civitai public API (base model MOSS Image), inclusion window 11-04 21:00 → 11-05 09:00 (Beijing Time), backfill period. Recommendations are reviewed by editors; download counts reflect attention only.