Zhenzhen's AI Radar · MOSS Image · 2026-11-16 Morning Report
This issue features 60 new versions: Strongly Recommended 1 · Recommended 15 · Worth Watching 19 · Listed Only 25
Editor's Note This issue reviews 61 new MOSS versions: 2 S-tier, 15 A-tier, 19 B-tier, and 25 C-tier. Top pick: Juggernaut Moss Turbo, which merges the author's self-trained DMD2 acceleration LoRA into the Juggernaut Moss realistic base model, generating images in 6 steps with CFG 1. The same author also released a standalone mountable DMD2 LoRA version. Another S-tier entry is the manga colorization LoRA MOSS Manga Colorist, which uses the anima_lora command-line reference image workflow with full parameters provided. Shrekman's batch of 90s animation style transfers, FRAME clothing details, and Nissan 180SX car models are listed as A-tier. The Emilia character collection added 9 outfit versions at once, all listed as C-tier; Prywinko's base model field was labeled as Anima, but after verifying the version name and file, it is retained as MOSS and listed as C-tier.
Strongly Recommended (1)
[S] MOSS Manga Colorist | Cinematic Manga Colorization · v1.0
LoRA · New Model · Author: moonrailgun · Published 11-16 02:27 · Downloads 52
MOSS Manga Colorist: A MOSS-Image Base LoRA for coloring black-and-white manga pages. Input a manga original page as a reference image to output colored pages with cinematic lighting, warm skin tones, and anime-style rendering.
Images

Reasons for Recommendation
- New model; serves manga colorization and black-and-white to color conversion: It uses a reference image inference workflow (original image as ref_image), not standard text-to-image. The author explicitly states that generic LoRA loaders cannot reproduce this effect.
- Actionable: The author provided complete command lines, inference.toml in the Usage ZIP, and example settings (prompt: comic, strength 1.0, 30 steps, CFG 3, seed 7, ER-SDE).
- 9 visible examples. The author states the gallery is a three-column comparison of a five-page manga (original / base model without LoRA / LoRA first pass), and proactively explains that the two columns use different inference workflows, so it is not a strict on/off comparison. Weights are Apache 2.0, and the Civitai license allows for image generation and sales.
How to Use
- Install anima_lora and download its MOSS Base model package (GUI or make download-model ARGS=moss-base); place the weights and the extracted Usage ZIP into the manga-colorist folder within the anima_lora directory.
- Run uv run --no-sync python scripts/moss/generate.py --config manga-colorist/inference.toml --ref_image <your manga page> --out_dir <output directory> in the anima_lora directory. You can pass an entire folder for batch colorization.
- Prompt: comic, strength 1.0, 30 steps, CFG 3, ER-SDE; keep other settings at inference.toml defaults; the base model must be MOSS-Image Base, and the LoRA cannot be used standalone.
Risks and Limitations
- Depends on anima_lora's command-line inference workflow; the author states that Civitai online generation and standalone ComfyUI workflows have not been verified, so ComfyUI users currently lack a ready-made integration.
- The author describes this as stylized colorization with repainting: faces, clothing, lines, and dialogue text may be altered, with no guarantee of preserving official character colors or original manuscript integrity.
- Copyright of example manga belongs to the original authors and publishers; the comparison images are not controlled variable experiments, so the LoRA's actual benefit must be judged individually.
Trigger Words: comic
Recommended (15)
[A] Dark Fantasy 90's anime style Moss/Krea2/Anima/illustriousXL/ZImageTurbo/Flux/Chroma · Moss
LoRA · Port · Author: zura_janai · Published 11-16 05:32 · Downloads 44
MOSS version of Dark Fantasy 90's anime style: A style LoRA overlaying 90s retro animation aesthetics onto dark fantasy themes.
Images

Reasons for Recommendation
- Port: The same page already includes Krea 2, Anima, Illustrious, Z-Image Turbo, Flux.1 D, and Chroma versions; the MOSS version differs from the Krea 2 version only in that the trigger words omit 'film grain', and the filename indicates training on MOSS-Image SFT.
- Serves fantasy character and scene illustrations in a retro animation style; the author states training images were sourced from glitchorange on Instagram.
- Parameters and trigger words are fully provided: strength 0.8–1, with 1 being optimal; 12 visible examples (8 hidden); free, license allows for image generation and rental; the entire series has high attention, with the MOSS version receiving 44 downloads in 3.5 hours after release.
How to Use
- Add dark_fantasy_style, dark fantasy, retro to the prompt, LoRA strength 1 (range 0.8–1).
- Steps, CFG, and sampler are not provided by the author; the filename includes 'sft', so start with an SFT base model.
Risks and Limitations
- Training materials are taken from an Instagram creator's work; the author does not disclose material licensing, so commercial use requires self-assessment.
- The description only lists parameters and does not detail the effect differences between the MOSS version and other base model versions; the page contains mixed content.
Trigger Words: dark_fantasy_style, dark fantasy, retro
[A] Juggernaut Moss DMD2 Lora · v1.0 Beta
LoRA · New Model · Author: KandooAI · Published 11-16 05:40 · Downloads 58
Juggernaut Moss DMD2 Lora (v1.0 Beta): A DMD2 distillation acceleration LoRA trained by the author specifically for their Juggernaut Moss. When mounted, it generates images in 6 steps with CFG 1, replacing the standard 30 steps and CFG 3.
Images

Reasons for Recommendation
- New model, released in tandem with the same-day Juggernaut Moss Turbo: Turbo is the integrated version with this LoRA pre-merged into the base model; this is a standalone LoRA that works best mounted on Juggernaut Moss V1, and the author states it also works on the original MOSS-Image Base.
- The author states they did not directly use existing MOSS acceleration LoRAs but trained this one to fit Juggernaut's style, achieving approximately five times the speed, with visuals close to but not identical to the full version.
- Node connections and parameters are fully documented, with a ComfyUI workflow available for direct drag-and-drop on HuggingFace; 14 visible examples (6 hidden); free, license allows for image generation and sales.
How to Use
- Place the LoRA in ComfyUI/models/loras/, and add a LoraLoaderModelOnly node between Load Diffusion Model (loading Juggernaut Moss) and the sampler, with strength 1.0 (0.9 is also acceptable).
- 6 steps (use 7 steps for complex multi-person or hand actions), CFG 1, euler_ancestral + simple; leave the negative prompt empty as it has no effect at CFG 1; resolution approx. 2 MP; the author recommends pairing this Beta with Juggernaut Moss's int8 files.
- The author states dpmpp_sde + beta yields similar results but is twice as slow; avoid karras (causes blurriness) and dpmpp_2m with normal or beta (causes artifacts); any text appearing in the image must be written in quotes, and text must also be provided for natural text locations like signs and labels, otherwise garbled text will appear.
Risks and Limitations
- The author describes this as a first-attempt Beta: busy scenes occasionally have extra fingers or strange arms, seed-to-seed variation is smaller than the full version, and composition tends to be centered; a new version trained on BF16 is in progress.
- The Turbo integrated version already includes this LoRA, so stacking them is redundant; the author only stated it works on Base, without providing performance data on SFT or other MOSS fine-tuned base models.
- The filename says 4-8Steps, but the author recommends 6–7 steps; performance with fewer steps is not specified; the page contains mixed content.
[A] Dirty-Moss · Dirty-MossV2
Base Model / Checkpoint · Version Update · Author: Sateluco · Published 11-16 00:58 · Downloads 1041
Dirty-Moss V2: A realistic photography integrated base model aiming for 'candid real-life snapshots', deliberately avoiding the over-polished look common in AI images, favoring daily scenes, natural skin, imperfect lighting, and casual composition.
Images

Reasons for Recommendation
- Version update: The author states at the outset that V2 is not necessarily better than V1, just different, and both versions are retained; the author claims V2 looks less like a polished image, has more natural expressions, different 'imperfect' handling of textures and lighting, and is overall rougher and less idealized, with the most noticeable differences in daily scenes and portraits.
- Parameters adjusted per version: CFG 2.5, 25 steps, Shift 5, ER SDE + Simple (Euler + Beta can also be tried); it remains a general realistic model capable of generating portraits, daily photography, and various environments.
- Provides both bf16 full precision and CONVROT8 (int8 quantization, significantly lower VRAM requirements) files; free, license allows for image generation, rental, and sales.
- Approximately 1041 downloads in 8 hours after release, and 15762 downloads for the entire page, making it the most attention-grabbing entry in this group; download counts only indicate attention level.
How to Use
- Load as a MOSS base model, no trigger words; write prompts focusing on daily snapshots, ordinary environments, and natural light, which aligns best with the author's stated strengths.
- CFG 2.5, 25 steps, Shift 5, sampling ER SDE + Simple; the author also provided preferred resolution and Hires.fix settings (Lanczos upscaling, low denoising strength), see the Risks section for specific values.
Risks and Limitations
- All improvements are author-claimed, and the author themselves states V2 is not necessarily better; 5 visible examples, 15 hidden, page contains mixed content.
- Parameters only found in the full model card: preferred resolution 1152*1440 (maximum recommended up to 2024x2024, as per original text), Hires.fix uses Lanczos, 8 steps, denoising 0.15, upscale 1.5 times; text encoder and VAE are not mentioned by the author, bf16 file is approx. 10GB.
【A】90's anime melancholy Moss/Krea2/Anima/illustriousXL/ZImageTurbo/Flux/Chroma · Moss
LoRA · Port · Author zura_janai · Published 11-16 05:23 · Downloads 45
MOSS version of 90's anime melancholy: A retro-style LoRA focusing on the melancholic and nostalgic atmosphere of 90-era animation.
Images

Reasons to Recommend
- Port: The same page already features Krea 2, Anima, Illustrious, Z-Image Turbo, Flux.1 D, and Chroma versions; same author and template as Dark Fantasy 90's; the MOSS version's trigger words and strength are identical to the Krea 2 / Anima versions
- Serves emotional character and scene illustrations in a retro animation style
- Complete parameters: Strength 0.8–1, 1 is optimal; 8 visible examples (4 hidden); free, license allows image generation and renting; the entire series has high attention, with the MOSS version receiving 45 downloads in 3.6 hours since release
How to Use
- Add 90s_melancholy, retro to the prompt, LoRA strength 1 (range 0.8–1)
- Steps, CFG, and sampler are not provided by the author; the filename includes 'sft', so start with an SFT base model
Risks and Limitations
- The filename indicates epoch-01, but the author does not specify if this is the final epoch, making it impossible to judge if it is undertrained
- The description only lists parameters, without mentioning the source of training materials or the performance differences of the MOSS version; the page contains mixed content
Trigger Words: 90s_melancholy, retro
【A】Retro Neo Noir Moss/Krea2/Anima/illustriousXL/ZImageTurbo/Flux/Chroma · Moss
LoRA · Port · Author zura_janai · Published 11-16 05:45 · Downloads 43
MOSS version of Retro Neo Noir: A style LoRA combining 90-era retro animation aesthetics with a neo-noir atmosphere.
Images

Reasons to Recommend
- Port: The same page already features Krea 2, Anima, Illustrious, Z-Image Turbo, Flux.1 D, and Chroma versions; same author and template, the MOSS version's trigger words and strength are the same as most versions
- Serves retro animation-style illustrations for urban night scenes, detective, and crime themes
- Complete parameters: Three trigger words, strength 0.8–1, 1 is optimal; 6 visible examples (11 hidden); free, license allows image generation and renting; the MOSS version received 43 downloads in 3.3 hours since release
How to Use
- Add retro_neo_noir, retro, retro artstyle to the prompt, LoRA strength 1 (range 0.8–1)
- Steps, CFG, and sampler are not provided by the author; the filename includes 'sft', so start with an SFT base model
Risks and Limitations
- Hidden examples outnumber visible examples; the description only lists parameters, without mentioning the source of training materials or the performance differences of the MOSS version
- The page contains mixed content
Trigger Words: retro_neo_noir, retro, retro artstyle
【A】Ink Noir — Graphic Novel Woodcut Style · v1.0
LoRA · New Model · Author kentus · Published 11-16 02:09 · Downloads 48
Ink Noir: A hard-edged black-and-white woodcut / printmaking comic style LoRA, inspired by early 20th-century woodcuts, end-grain woodcuts, and lithographs—featuring large solid blacks, carved white lines, silhouettes, and strong night lighting; it can also depict modern subjects like mobile phones, cars, neon signs, and subways.
Images

Reasons to Recommend
- New model; serves wordless graphic novels, film noir-style posters, night scenes, and silhouette illustrations; the author states the trigger word itself carries the style, so there is no need to write ink, woodcut, or black and white in the prompt
- Specific usage instructions: Weight 0.8–1.0, lower to 0.5–0.7 to preserve more model detail; best suited for night streets, indoor scenes under lamps, silhouettes, rain scenes, crowds, and hard side-lit portraits
- 6 visible examples (2 hidden); free, license allows image generation and selling
How to Use
- Start the prompt with the trigger word, then describe the scene in plain language; the description lists 'inknoir', but the registered trigger word is 'inknoire', try both
- Weight 0.8–1.0 (0.5–0.7 preserves more detail); for deeper blacks add night, single lamp, or deep shadows; for finer carving lines add detailed or wide shot
- Steps, CFG, and sampler are not provided by the author
Risks and Limitations
- Inconsistent spelling between the description and the registered trigger word (inknoir / inknoire), unexplained by the author
- Version is labeled MVP v1.0; whether to use Base or SFT base model, and sampling parameters are not specified; the page contains mixed content
Trigger Words: inknoire
【A】FRAME Fashion Fidelity · MOSS v1.0
LoRA · Port · Author Coffmeteor · Published 11-16 01:19 · Downloads 41
MOSS version of FRAME Fashion Fidelity: A garment fidelity LoRA that makes character outfits closer to fashion reference images, primarily helping to render complex clothing details (side slits, layered knits, lace trim, pleats) correctly, ensuring full-body lookbook compositions do not cut off heads, and making poses and expressions more like street photography.
Images

Reasons to Recommend
- Port: The same page previously featured a Krea 2 version; the MOSS version filename is Garment Reconstruction (the Krea 2 version was called Expanded Garment Reconstruction), indicating the description follows the Krea 2 version, with no MOSS-specific differences written by the author
- Serves fashion e-commerce images, outfit lookbooks, and fashion portraits: The author claims it pushes garments, expressions, and body proportions toward the reference aesthetic used by FRAME, while admitting they cannot clearly explain the specific mechanism
- Many examples: 10 images and 2 videos visible (4 hidden); free, license allows image generation, selling, and derivatives, with no attribution required
How to Use
- No trigger word; after loading the LoRA, write normal prompts for clothing and characters; the author suggests a strength of 1.0
- Steps, CFG, sampler, and whether to use Base or SFT base model are not provided by the author
Risks and Limitations
- The original text for the strength 1.0 recommendation refers to Krea 2 (official Turbo and standard versions); it is not specified if this is equally suitable for MOSS
- The author admits they cannot clearly explain what it actually changes, and effect descriptions are all 'tends to', with no before-and-after comparison; the page contains mixed content
【A】Nissan 180SX (S13) Car LoRA · v1.0
LoRA · New Model · Author redboutique · Published 11-16 02:35 · Downloads 26
Nissan 180SX (S13) car model LoRA: Renders this Nissan sports car on MOSS, allowing customization of paint, wheels, wide-body kits, spoilers, pop-up headlight states, and scenes.
Images

Reasons to Recommend
- New model; serves automotive photography-style image generation, modified car concept art, and anime livery (itasha) design
- The author states the training set consists of 206 individually captioned images, covering stock styles, lowered street cars, large spoilers, colored wheels, full-body anime liveries, as well as underglow and backfire exhaust flames; pop-up headlights have states for retracted, raised, half-open, and one-eye wink
- Provides trigger words and a full copyable example prompt; 11 visible examples (9 hidden); free, license allows image generation and selling
How to Use
- Start the prompt with N180_SX, then describe paint, wheels, kits, spoilers, headlight states, and scenes, e.g., the author's example: N180_SX, a metallic midnight-blue Nissan 180SX with wide fenders, white six-spoke wheels, a black GT wing and raised pop-up headlights…
- Strength, steps, CFG, and sampler are not provided by the author
Risks and Limitations
- Covers only this specific car model, narrow use case; no sampling parameters provided
- Real car model and brand name; commercial image generation requires attention to trademarks; the page contains mixed content
Trigger Words: N180_SX
【A】Energetic Funk · v1.0 Moss
LoRA · Port · Author Daalis · Published 11-15 23:03 · Downloads 31
MOSS version of Energetic Funk: A style LoRA featuring high-saturation colors, strong graphic design feel, exaggerated expressions, and stylized proportions, blending the exaggeration of Japanese animation and Western cartoons.
Images

Reasons to Recommend
- Port: The same page already features Krea 2, Anima, Z-Image Turbo, Illustrious, Flux.2 Klein 9B-base, and Flux.1 D versions; the MOSS version description states this is a batch where the author is porting old styles to MOSS one by one, claiming this style performs well on MOSS, with a personal preference strength of about 0.8, lower than the general recommendation of 1.0
- Serves character standing illustrations, full-body stylized characters, chibi monsters and companions, trendy clothing, and colorful fantasy designs
- Complete parameters: Trigger word FunkDaal, recommended strength 0.5 to 1.5, style lock is strongest at 1.0; the author states example images include prompts and workflows; free, license allows image generation
How to Use
- Add FunkDaal to the prompt, then describe the character, clothing, mood, and environment; on MOSS, start from 0.8, increase above 1.0 for a stronger style
- Steps, CFG, and sampler are not written in the description by the author, who states they can be viewed in the generation info of the example images
Risks and Limitations
- The author self-describes this style as 'hot leaning' (tending towards sexualization), the page contains mixed content, with 3 visible examples and 3 hidden
- The author does not describe the performance differences between the MOSS version and other base model versions; the license only allows image generation and on-site renting
Trigger Words: FunkDaal
【A】Monster Art illustration [Krea + MOSS + Flux1/2 + ZiT + IL] · MOSS
LoRA · Port · Author ColonelJay · Published 11-16 05:40 · Downloads 16
MOSS version of Monster Art illustration: A style LoRA featuring scratchboard-style dense carving lines and cross-hatching, distorted and elongated bodies, strong chiaroscuro, and high-density details, leaning towards dark, mysterious, and psychedelic fantasy illustration.
Images

Reasons to Recommend
- Port: The same page already features Krea 2, Flux.2 Klein 4B, Flux.1 D, Z-Image Turbo, and Illustrious versions; the MOSS version description only says 'MOSS image version', with no differences written
- Serves card illustrations, tabletop RPGs and game art, dark fantasy and sci-fi illustrations; the author breaks the style down into four points: carving texture, organic distortion, high-contrast atmospheric lighting, and layered details, which can be supplemented in prompts
- The author claims all examples were generated using only the base model plus this single LoRA; 10 visible examples (10 hidden); free, license allows image generation
How to Use
- The MOSS version has no registered trigger word (monster_4rt_illustr4tion is for the Illustrious version); add the author's provided sentence to the prompt: An illustration in a scratchboard art and cross-hatching style. or This is a digital illustration in a scratchboard art and cross-hatching style.
- LoRA strength 1; steps, CFG, and sampler were not provided by the author
Risks and Limitations
- The author cites living illustrators such as Eric Sabee as style references, which may raise commercial style imitation disputes
- MOSS-specific parameters and differences from other versions are not documented; the page contains mixed content
【A】Blue Memphis · v1.0-moss-heal
LoRA · Port · Author soda_Cookie · Published 11-16 00:51 · Downloads 33
The MOSS version of the Blue Memphis art style LoRA: features bright blue and cyan as primary colors, accented with coral orange and cream white. Characters, clothing, backgrounds, and compositions are all pulled into this color scheme, with lighting simplified into blocks and distinct color zoning.
Image

Reasons to Recommend
- Port: The Anima version was released first on the same day; the MOSS version is named v1.0-moss-heal, and the file is epoch 36 of v002. The author did not explain the meaning of 'heal' nor write about differences from the Anima version
- Serves unified color-stylized illustrations, posters, and character standing art; the author provided positive and negative prompts for a complete example (blue flower shop scene, 832×1216, seed 715501) that can be directly copied or modified
- 4 visible examples; free, license allows image generation and sale
How to Use
- Write In @memphis style. at the end of the prompt, and specify the blue main tone with small amounts of coral orange and cream white accents in the body text as per the author's example; use worst quality, low quality, blurry, text, watermark, etc. for negatives
- Strength, steps, CFG, and sampler were not provided by the author
Risks and Limitations
- The model page description refers to v002 EP12 (corresponding to the Anima version file), with no specific instructions for the MOSS version
- The author's example prompts already specify the color scheme; it is unclear how strong the style is when only the trigger word is used; there are only 4 examples
Trigger Word: In @memphis style.
【A】[MOSS] Riding Chocobo · v1
LoRA · New Model · Author NanashiAnon · Published 11-16 08:05 · Downloads 6
[MOSS] Riding Chocobo: Allows characters to ride Chocobos from the Final Fantasy series, with options to specify the bird's color, saddle, stirrups, and reins.
Image

Reasons to Recommend
- New model (concept LoRA); serves FF fan art and fantasy mount scenarios: solves the issue where base models often fail to correctly render the dual-subject composition of 'a person riding a large bird'
- Tagging syntax is specific: must include R_C_C_L, riding, [color] chocobo, saddle; optional stirrups (riding), holding reins (add two-handed if holding reins with both hands), barding; explicitly states not to write solo, as the bird and rider are two subjects
- The author claims it uses the same training data as their Illustrious version; 4 visible examples, content rating is safe; free, license allows image generation, rental, and sale
How to Use
- Write R_C_C_L, riding, yellow chocobo, saddle in the prompt (color can be changed), add stirrups (riding), holding reins, barding as needed; do not add solo
- If holding reins with only one hand, clearly specify what the other hand is doing; strength, steps, and CFG were not provided by the author
Risks and Limitations
- Published less than 1 hours ago, with few downloads and feedback; strength and sampling parameters were not provided by the author
- Chocobo is a copyrighted image of Square Enix; commercial use requires self-assessment
Trigger Words: R_C_C_L, riding, yellow chocobo, saddle,, stirrups (riding), holding reins, barding
【A】Mimics: Friendly Objects of Mischief for MOSS · v1.0
Base Model / Checkpoint · New Model · Author Razira · Published 11-16 08:56 · Downloads 1
Mimics: Friendly Objects of Mischief: Transforms any object into a 'mimic' (a biting object like a chest mimic)—adding grinning sharp teeth, a long tongue, and randomly placed eyes to objects.
Image

Reasons to Recommend
- New model (site type labeled as Checkpoint, but the author's description states it is a LoRA, and the file size is LoRA-level); serves fantasy tabletop RPGs, monster design, and gag prop concept art
- The author provided the trigger word m1m1c aesthetic and a modifiable monster description: Monstrous grinning mouth with sharp teeth and a long tongue. It has many randomly placed eyes
- The author claims MOSS can do this without it, but this LoRA helps significantly; it brings a slight illustration concept design style, but it is weak and can be overridden by prompts or other LoRAs; 8 visible examples (11 hidden)
How to Use
- Write m1m1c aesthetic (or m1m1c style) in the prompt, then describe the object and its mouth, teeth, tongue, and eyes; do not just write m1m1c, as the author claims this will randomly print the text 'm1m1c' on the image
- Load as a LoRA, do not place in diffusion_models as a base model; strength, steps, and CFG were not provided by the author
Risks and Limitations
- The author notes that effects vary by image; strength and sampling parameters are missing
- Site type incorrectly labeled as Checkpoint; license only allows on-site rental, not derivatives; recently published, page contains mixed content
Trigger Word: m1m1c aesthetic
【A】moss base flipping the bird · v1
LoRA · New Model · Author hamstornado · Published 11-16 00:22 · Downloads 89
moss base flipping the bird: A concept LoRA for the middle finger gesture, supplementing MOSS's rendering of this specific gesture.
Image

Reasons to Recommend
- New model; serves memes, gag posters, and rebellious character illustrations requiring an accurate middle finger gesture, as hand poses are usually difficult to render correctly with prompts alone
- The author provided two methods: trigger word fl1pp1ng-b1rd, or natural language she / he is making a crude middle finger hand gesture
- Approximately 9 hours after publication, it has 89 downloads, the highest among new models in this group; 2 visible examples (4 hidden), free, license allows image generation
How to Use
- Write fl1pp1ng-b1rd in the prompt, followed by she is making a crude middle finger hand gesture (change to he for males)
- Strength, steps, and CFG were not provided by the author; the name indicates moss base, compatibility with SFT is not specified
Risks and Limitations
- The description contains only two lines of prompts, with no training details or parameters
- Single-purpose, handling only this gesture; images with vulgar gestures may be restricted on some platforms; page contains mixed content
Trigger Word: fl1pp1ng-b1rd she is making a crude middle finger hand gesture
【A】Fwaun Tune · moss_image_v5.2
Base Model / Checkpoint · Version Update · Author fwaunstp733 · Published 11-16 07:16 · Downloads 6
Fwaun Tune's MOSS v5.2 full fine-tuning base model: Specializes in flat coloring, lineless art, and semi-chibi (not full chibi) distorted character art styles.
Image

Reasons to Recommend
- Version update: The model's general description changelog states MOSS v5.2 switched to fizgig training and the CAME optimizer; the previous MOSS v5.1 on the same page had its learning rate lowered due to overfitting causing poor LoRA stacking effects, and v5.0 was the first MOSS version
- Serves flat illustrations and lineless chibi character art; the author claims the Krea 2 / Anima versions in the same series can produce monochrome and realistic backgrounds
- Provides int8 (ConvRot) and bf16 weights, 3 visible examples; free, license allows image generation, rental, and derivatives, the entire model page already has many downloads
How to Use
- Load as the main MOSS model; the int8 version saves VRAM, the bf16 version has higher precision; v5.2 has no registered trigger word, previous MOSS versions registered fwaun style, semi-chibi style, which can be used initially
- Steps, CFG, and sampler were not provided by the author; dataset and training parameter details are on the author's Fanbox / Patreon
Risks and Limitations
- The version description section is empty, the author did not write update notes, only mentioning the switch in training tools and optimizer in the general description, without detailing the effect difference compared to v5.1
- The full base model is large; v5.2 was published about 2 hours ago with few downloads, 1 examples are hidden
Worth Watching (19)
【B】Rookie - Halo ODST · Rookie_ODST_Moss_v1.0
LoRA · Port · Author AthosAI · Published 11-16 06:06 · Downloads 10
MOSS character LoRA for Rookie, the protagonist of Halo 3: ODST: features a dark gray combat helmet with light gray thick stripes, a matte blue-gray visor, and optional full ODST heavy armor.
Image

【B】Dark Art Style Moss/Krea2/Anima/illustriousXL/ZImageTurbo/Flux/Chroma · Moss
LoRA · Port · Author zura_janai · Published 11-16 05:15 · Downloads 58
MOSS version of the Dark Art Style LoRA: The author's 'Dark Art Style' series across seven base models adds a MOSS version, with the description containing only the trigger word and weight, without specifying what this style looks like.
Image

【B】Cat Matroskin (Prostokvashino) · v1.0 MOSS
LoRA · New Model · Author bazya022020810 · Published 11-16 05:41 · Downloads 14
Character LoRA for Matroskin the cat from the classic Soviet animation Prostokvashino, rendered with a plush toy and live-action film texture: gray-blue striped fur, yellow eyes, pink nose, and a red-green striped fringed scarf.
Image

【B】90s anime aesthetic Moss/Krea2/Anima/illustriousXL/ZImageTurbo/Flux/Chroma · Moss
LoRA · Port · Author zura_janai · Published 11-16 05:39 · Downloads 54
MOSS version of 90s anime aesthetic: A LoRA for the Japanese anime art style of the 90s, the most versatile entry in the author's retro animation series.
Image

【B】DRKM4GE — Chiaroscuro · v1.0-MI
LoRA · Port · Author theevildarkmage126 · Published 11-15 21:22 · Downloads 52
DRKM4GE — Chiaroscuro for MOSS: An artistic enhancement LoRA focused on dramatic chiaroscuro, dimensional shadows, and fine materials (metallic reflections, gems, glass, fabrics), originally trained and released on Krea 2.
Images

【B】Chiaroscuro Moss SFT · Int8_convrot v1.0
Base Model / Checkpoint · Version Update · Author svnx7x_ · Published 11-16 02:49 · Downloads 19
Int8 ConvRot quantized version of Chiaroscuro Moss SFT: The author merged two LoRAs (High Landscape, Character Detailer Booster) into MOSS-Image-SFT to create an integrated base model for anime-style images with enhanced color contrast and detail; this version compresses the file size to approximately half.
Images

【B】moss base loser hand gesture · v1
LoRA · New Model · Author hamstornado · Published 11-16 00:05 · Downloads 61
Loser hand gesture LoRA: Makes characters form an 'L' shape with their thumb and index finger (the 'loser' gesture); the position of the hand can be specified in the prompt.
Images

【B】Lonecat’s Simple Speed Boost 🚀 Workflows | 简易加速工作流 (MiniMax H3, Krea 2, Qwen 2.1, ZIT, Klein_9b, Illustrious, Flux, Wan2.2, Qwen Image Edit) · Simple MOSS Image
Workflow · Port · Author lonecatone23 · Published 11-15 21:21 · Downloads 60
MOSS version of Lonecat's Simple Workflow series: A ComfyUI text-to-image workflow with sectioned annotations for beginners, featuring built-in Relay VSR upscaling.
Images

【B】赤冬/akafuyu(明日方舟/arknights) · v1.0
LoRA · New Model · Author DSLM · Published 11-16 01:40 · Downloads 18
MOSS character LoRA for Akafuyu from Arknights: Features purple hair with red highlights, orange eyes, and a small mole below the lip; covers three outfits: Elite 0 default skin, Elite 2 default skin, and the 'Xinxing' (Heart Nature) skin.
Images

【B】Stockings Concept for Moss-Image and Krea-2 · Moss-Image
LoRA · Port · Author arkinson · Published 11-16 04:46 · Downloads 41
MOSS version of the Stockings Concept fashion LoRA: Generates more fashionable stockings and pantyhose styles, along with matching poses and outfits.
Images

【B】伺夜/vigil(明日方舟/arknights) · v1.0
LoRA · New Model · Author DSLM · Published 11-16 01:35 · Downloads 13
MOSS character LoRA for Vigil from Arknights: Features black hair, olive eyes, wolf ears and tail, and earrings; covers three outfits: Elite 0 default skin, Elite 2 default skin, and the 'Syracuse's Other Face' skin.
Images

【B】Bloodybeni Style [Anima/Krea2/MOSS] · MOSS v1.0
LoRA · Port · Author Imagine__That · Published 11-15 23:18 · Downloads 58
MOSS version of the Bloodybeni Style LoRA: A portrait illustration style with rich colors and visible painterly brushstrokes; the author claims it makes both male and female subjects look better.
Images

【B】Pencil Drawing Enhancer - CE · V01 - MossImage Base/SFT
LoRA · Port · Author CreativeEdge · Published 11-16 01:29 · Downloads 44
MOSS version of Pencil Drawing Enhancer: Renders images with a colored hand-drawn feel featuring distinct pencil lines, in a colored pencil illustration style.
Images

【B】Opulent Darkness · MOSS Base
LoRA · Version Update · Author ArsMachina · Published 11-15 22:23 · Downloads 22
MOSS Base version of Opulent Darkness: A LoRA replicating the author's personal Midjourney 8.2 style, featuring luxurious dark fantasy TCG card illustration aesthetics—gold-inlaid floral patterns, Baroque decorations, and iridescent bioluminescent highlights. This file was trained on MOSS-Image Base and is available for free download.
Images

【B】Vyshivanka вишиванка (MOSS) [Anime & Realism] · v1.0
LoRA · New Model · Author EDG · Published 11-16 00:00 · Downloads 32
MOSS version of the Vyshivanka (вишиванка) clothing LoRA: Dresses characters in traditional Ukrainian embroidered shirts; the title states it works for both anime and realistic styles.
Images

【B】Dark Ghibli Fairytales · MOSS
LoRA · Port · Author ArsMachina · Published 11-15 22:22 · Downloads 5 · Paid
MOSS version of the Dark Ghibli Fairytales style LoRA: Combines Ghibli-style animation aesthetics with dark fairy tales—watercolor backgrounds, cel-shaded characters, ethereal lighting, and fantastical natural environments, offering a gentle yet slightly eerie atmosphere.
Images

【B】ColoredPencil [Minimax / MOSS / Krea2 / SD] · MOSS v1.0
Base Model / Checkpoint · Port · Author TeraWatt · Published 11-16 04:28 · Downloads 0 · Paid
MOSS version of ColoredPencil: Renders images as traditional colored pencil drawings—soft strokes, layered coloring, and paper texture, leaning towards a children's book and hand-drawn illustration aesthetic.
Images

【B】moss base sph hand gesture · v1
LoRA · New Model · Author hamstornado · Published 11-16 00:35 · Downloads 0 · Paid
moss base sph hand gesture: A LoRA for the 'small' hand gesture concept, where the thumb and index finger pinch together to indicate a tiny distance; by the same author as the middle finger LoRA.
Images

【B】Texture Enhancer - Micro-Contrast for Moss · v1.0
LoRA · New Model · Author Liant · Published 11-16 00:26 · Downloads 0 · Paid
Texture Enhancer - Micro-Contrast: A general-purpose LoRA that enhances local micro-contrast, fine surface textures, and shadow depth, making details in skin, fabric, hair, and backgrounds clearer and more volumetric.
Images

Included Only (25)
- Emilia (re:zero) {AIO - seperate} [MOSS] · Season3/wedding — LoRA: The third season / wedding version of the Emilia outfit collection: a white off-the-shoulder long dress, two-tone necklace, twin braids, and white long gloves.

- Moss Image AI Babe Pack ( Part 04 ) by Sarcastic TOFU · MI_Babe_Rafaela_v1 — LoRA: A fictional female character LoRA from Part 4 of Sarcastic TOFU's "MOSS AI Babe Pack": Rafaela Noite, a Brazilian woman with golden-brown skin, dark brown curly hair, and green eyes.

- Moss Image AI Babe Pack ( Part 03 ) by Sarcastic TOFU · I_Babe_Niravi_v1. — LoRA: A character LoRA for the fictional female character Niravi (Sri Lankan descent, golden-brown skin, long dark hair, amber eyes) from Part 3 of Sarcastic TOFU's "Moss Image AI Babe Pack".

- Moss Image AI Babe Pack ( Part 01 ) by Sarcastic TOFU · MI_Babe_Malaika_v1 — LoRA: Malaika from the first part of Sarcastic TOFU's MOSS fictional female character pack: a Kenyan-American female character with black afro hair, dark skin, and brown eyes.

- Sneer [Krea 2/MOSS Image] · MOSS — LoRA: Sneer's MOSS version expression LoRA: makes the character's upper lip curl up, creating a teasing sneer expression.

- MOSS-Lu Xueqi · MOSS-Lu Xueqi — LoRA: A MOSS character LoRA for Lu Xueqi from "Jade Dynasty"; the author provides close-up portrait prompts with BJD doll material and 3D rendering texture.

- Moss Image AI Babe Pack ( Part 02 ) by Sarcastic TOFU · MI_Babe_Leilani_v1 — LoRA: A character LoRA for the fictional female character Leilani (Hawaiian / Polynesian descent, golden-brown skin, long dark brown hair, dark brown eyes) from Part 2 of Sarcastic TOFU's "Moss Image AI Babe Pack".

- Emilia (re:zero) {AIO - seperate} [MOSS] · Date — LoRA: The Date version of the Emilia outfit collection: a white dress with a pink corset, puff long sleeves, and a pink bow, wearing an emerald necklace.

- Emilia (re:zero) {AIO - seperate} [MOSS] · Trial/Closed — LoRA: The Trial / Closed version of the Emilia outfit collection: a grey collared dress with a purple scarf and purple accents, wide long sleeves, and an emerald necklace.

- TheMinusculeTask Style [Anima + MOSS] · MOSS v1.0 — LoRA: The MOSS version LoRA of artist The-Minuscule-Task's style: trained only on flat-color works, leaning towards beastman / anthropomorphic animal characters.

- Nazuna Hiwatashi (Brand New Animal, BNA) · v2.0 — LoRA: MOSS character LoRA v2.0 for the anime "BNA" character Nazuna Hiwatashi: a fox beastman girl with pink fur.

- Emilia (re:zero) {AIO - seperate} [MOSS] · Season4 — LoRA: The fourth season version of the Emilia outfit collection: a purple bodysuit covered by a white round-edged cloak, featuring a cat-ear hood, purple cuffs, detached sleeves, and an open-back design.

- Emilia (re:zero) {AIO - seperate} [MOSS] · Pink/Morning — LoRA: The Pink / Morning version of the Emilia outfit collection: a pink casual dress with a red flower on the chest, cutouts on the shoulders, long sleeves, paired with white thigh-high socks.

- Emilia (re:zero) {AIO - seperate} [MOSS] · Drunk/MemorySnow — LoRA: The Drunk / MemorySnow version of the Emilia outfit collection: a light purple ruffled nightgown with an open dark purple jacket draped over it, exposing the shoulders.

- SoulTide 灵魂潮汐 Louisa 路易莎 DiT · MOSSTagFlowV1.1 — LoRA: A MOSS version character LoRA for Louisa (a dog-eared girl) from the game "Soul Tide"; it can switch between two outfits: a police uniform and a white apron, with an additional pure character prompt without clothing.

- Spakka5 Style · MOSS V1 — LoRA: The MOSS version LoRA of Spakka5's art style: a 2D illustration style with realistic shading, with tags pointing to Western cartoons and curvy body types.

- Emilia (re:zero) {AIO - seperate} [MOSS] · Capella — LoRA: The Capella version of the Emilia outfit collection: a plain white midi dress with triangular cutouts and shoulder cutouts, and detached sleeves.

- SoulTide 灵魂潮汐 Aria 阿莉娅 DiT · MOSSTagFlowV1.1 — LoRA: A MOSS character LoRA for Aria from the game "Soul Tide": multi-colored long hair in white, blue, and grey, with heterochromia (blue and purple eyes); it can switch between two outfits: a maid outfit and a Valentine's Day red bow dress.

- Michiru Kagemori (Brand New Animal, BNA) · v1.0 — LoRA: A MOSS character LoRA for the anime "BNA: Brand New Animal" character Michiru Kagemori, an anthropomorphic tanuki girl.

- Emilia (re:zero) {AIO - seperate} [MOSS] · LiMSwimsuit — LoRA: The LiM swimsuit version of the Emilia outfit collection: a multi-colored one-piece swimsuit with a black choker and purple brooch; the hairstyle is changed to black hair with a high ponytail and multiple hair flowers.

- Emilia (re:zero) {AIO - seperate} [MOSS] · Base — LoRA: The Base version of the Emilia outfit-specific LoRA collection from "Re:Zero - Starting Life in Another World": her iconic white and purple corset top with a layered pleated skirt, detached wide sleeves, thigh-high boots, and an emerald necklace.

- Opulent Darkness · MOSS — LoRA: The MOSS version of Opulent Darkness: the same ornate dark fantasy TCG style LoRA as the subsequently released MOSS Base version, but this version's file is trained with MOSS-Image SFT and is currently a paid download.

- Theodora (Fire Emblem: Fortune's Weave) [MOSS Image] · v1.0 — LoRA: A MOSS character LoRA for Theodora from the new "Fire Emblem" game Fortune's Weave, trained on MOSS Image Base using the Civitai trainer.

- Prywinko - Style - MOSS Image · MOSS Image — LoRA: A MOSS LoRA of artist prywinko's art style (the Civitai base model is mislabeled as Anima): tags point to 2D beautiful girl illustrations, noting that there is only one danbooru link.

- Very Simple txt2img MOSS Workflow · v1.0 — Workflow: Very Simple txt2img MOSS Workflow: a MOSS text-to-image ComfyUI workflow, with the file named MOSS_Artistic_Quality_V1, and the description contains only one performance reminder.

Data: Civitai public API (base model MOSS Image), inclusion window 11-15 20:31 → 11-16 09:01 (Beijing Time). Recommendations are editor-reviewed; download counts only reflect attention.