Zhenzhen's AI Radar · MOSS Image · 2026-11-19 Evening Report
This issue features 49 new versions: Strongly Recommended 2 · Recommended 10 · Worth Watching 13 · Included Only 24
Editor's Note This issue reviewed 51 new MOSS versions, with 2 rated S, 10 rated A, 13 rated B, and 25 rated C; 1 were excluded. Top pick: Bulle d’Encre preview version, a European comic-style fine-tuned base model featuring ink lines and flat coloring in the style of European comics. It can generate comic pages with panels and speech bubbles, where text inside the bubbles is generated exactly as specified in the prompt. All parameters and dependencies are provided. Another S-rated item is the Detail Slider (negative values soften, positive values sharpen). The free version of Moody Moss Mix V5.5 has been changed to generate images in 15 steps. Dizzle's several celestial effect LoRAs, tin-type old photos, and realistic phone snapshots are rated A. A large number of standard character LoRAs and styles with only one-line descriptions or fully hidden examples were downgraded to C; the "MOSS Base" version of Octane Render has the same weight hash as the "MOSS" version on the same page and has been excluded.
Strongly Recommended (2)
[S] Bulle d'Encre · Preview
Base Model / Checkpoint · New Model · Author _HaDeS_ · Published 11-19 20:43 · Downloads 3
Preview version of the European comic-style MOSS fine-tuned base model: Ink lines + flat coloring / watercolor style in the manner of European comics. It can generate comic pages and album covers with panels, narration boxes, and speech bubbles (text written according to the prompt).
The full model card states that both the base model and the self-trained LoRA use the Apache license; this group has the clearest attribution of sources.
Images

Reasons for Recommendation
- New capability: Instead of attaching a style LoRA, the author has baked their self-trained style LoRA and half-strength Turbo into the base model, allowing image generation in just 15 steps. The author claims that speech bubbles, cover titles, and small text on clothing are generated exactly as written in your prompt.
- Serves comic storyboard pages, retro adventure / film noir / spy-style covers, and black-and-white comic pages. Illustrious / NoobAI users can directly use Danbooru tags; long natural language descriptions also work.
- Parameters, text encoder, VAE, and negative prompts are all provided. The 19 example images uniformly use seed 1200 and the same settings, showing no sensitivity to seeds. Prompts are included in the image metadata. Free, with two files available: bf16 and int8.
How to Use
- Replace the base model for loading. Text encoder: qwen3.5_9b_int8_convrot (Load CLIP, select type moss_image). VAE: flux2-vae. Both are from Comfy-Org/MOSS-Image.
- 15 steps, CFG 2.5 (up to 5 is also acceptable), Euler / sgm_uniform (simple is also fine), flow shift 5, resolution 1344 × 1792 or above 2MP. Negative prompt: worst quality, bad anatomy, jpeg artifact.
- No trigger word is needed, and there is no need to attach Turbo again. For faster generation, you can stack the author's Turbo 8-steps v1.0 at strength 0.5, using 4-8 steps; the author notes that image quality will suffer.
- Text: Write the exact text to be displayed in the prompt, one bubble per line, and specify whose bubble it is and where it is located. Camera angles must be written as complete sentences; short tags are often ignored.
Risks and Limitations
- This is a preview version. The author states that the formal fine-tuning is still being trained on a larger dataset. The training images for the style LoRA are outputs from the author's own model.
- Known issues listed by the author: Interactions between two people often require correction or seed changes; brand logos may appear on everyday objects; occasional fake signatures, extra objects, or garbled small text in the background. It will generate adult content as requested.
- The full model card states that bf16 is approximately 10.2 GB and int8 is approximately 5.4 GB. The int8 version requires a newer version of ComfyUI, and the example images cannot be reproduced pixel-for-pixel. Optimal results for the examples depend on the author's HaDeS MOSS Suite prompt enhancement node (free).
[S] MOSS Detail Slider · v1.0
LoRA · New Model · Author baowy · Published 11-19 15:01 · Downloads 104
Full-scene detail slider: Negative values soften the overall image and increase fog density; positive values sharpen metal scratches, leather textures, gravel, and fabric textures. It is not just skin smoothing.
Suitable for creating a three-panel comparison with the same seed at −5 / 0 / +5.
Images

Reasons for Recommendation
- New capability: Acts on the entire scene (skin, materials, environment, atmosphere). The author clearly describes the changes at −5 / 0 / +5 using a weight table and claims that face shape, expression, lighting direction, and composition remain largely unchanged.
- Serves post-processing texture adjustment for realistic portraits and scene images: Pull towards negative values for a refined look; push towards positive values for a documentary-style gritty close-up. No need to change the base model.
- Daily range is −5 to +5. The author claims it does not break down even when pushed to ±10. No trigger word required. 104 downloads in 6 hours since release (indicating attention level only). Free, with a permissive license.
How to Use
- Load the LoRA; no trigger word needed. Daily strength is −5 to +5; ±10 is for special purposes.
- The author recommends pairing it with a 8-step DMD acceleration LoRA (strength 1.0).
- Steps, CFG, and sampler are not provided by the author.
Risks and Limitations
- Only 3 example images are available; there are no multi-step comparison images with the same seed. The effects of the weight table are as claimed by the author.
- Author's warning: High positive values may cause skin to appear reddish; negative values may cause faces to appear pale.
Recommended (10)
[A] Moody Moss 4KHD Workflow · v4.0 15 STEPS
Other · Version Update · Author catlover1937 · Published 11-19 16:58 · Downloads 115
Version 4 of the ComfyUI 4K HD workflow accompanying the Moody Moss series: Multi-stage generation plus upscaling, face refinement, and inpainting. V4 has been changed to support 15-step models for faster speed.
Images

Reasons for Recommendation
- Version update: V1–V3 were all labeled as 25 steps; the V4 version name has been changed to 15 STEPS. The model card entry 2026-11-19 states that it supports all Moody Moss base models and has been retuned for faster generation. When using other base models, you must enable the acceleration LoRA embedded in the workflow.
- Serves final finished images for realistic portraits: Modules listed by the author include multi-stage 4K upscaling, optional skin texture enhancement, face refinement with the ability to attach character LoRAs during the refinement stage, automatic and manual inpainting, optional SeedVR2 upscaling, and before/after comparison nodes. Each module can be toggled individually.
- 15 example images. The attachment is a 152 KB workflow JSON file containing no weights. 115 downloads in 4 hours since release (indicating attention level only).
How to Use
- Load Moody Moss 4k HD Workflow V4.json in ComfyUI. It is recommended to use the new version of Moody Moss Mix by the same author as the base model (also tuned for 15 steps).
- When using non-Moody Moss base models, enable the acceleration LoRA embedded in the workflow. SeedVR2 upscaling is an optional module.
- CFG, sampler, and required custom nodes are not specified by the author in the model card. The author states that Chinese and English instructions are included within the workflow.
Risks and Limitations
- The author claims it is much heavier than the Minimal version, requiring longer generation times and higher VRAM usage, especially when upscaling and refinement modules are enabled. Specific VRAM requirements are not provided.
- The V4 version description field is empty; changes are only noted in a single sentence on the model card. The list of required custom nodes and upscale models is not listed on the card.
- The license does not allow derivatives, and the author requires permission for reprinting. Additionally, 5 sensitive-level example images have been hidden.
[A] Renault 8 Gordini [Moss] · v1
LoRA · New Model · Author denrakeiw · Published 11-19 14:38 · Downloads 21
Single-model LoRA: Recreates the Renault 8 Gordini concept car (French blue body, offset white double stripes, six-light front end). It can generate both exterior views and retro rally-style cockpits.
Images

Reasons for Recommendation
- New model: The author trained on 60 manually selected and individually tagged high-resolution photos, covering studio shots from all angles, auto show scenes, exterior detail close-ups, and complete cockpit interiors.
- Serves automotive concept art, auto show posters, and product-render-style text-to-image generation. The model card lists details such as headlight clusters, rear diffusers, 19 / 20 inch mismatched front/rear wheels, and cockpit instruments item by item as usable prompts.
- Trigger word, strength, opening sentence structure, and 5 long example prompts are provided and can be copied directly. 17 visible examples. Free.
How to Use
- Trigger word: Renault 8 Gordini Concept. LoRA strength: 0.85 – 1.0.
- Author's recommended opening: "Create a photograph of a pearlescent French racing blue Renault 8 Gordini Concept ..." For cockpit shots, use: "Create a photograph of the driver-focused interior cockpit of a Renault 8 Gordini Concept ..."
- Steps, CFG, and sampler are not provided by the author.
Risks and Limitations
- It only learns one car model, so its use case is very narrow. All example prompts feature white studio or auto show scenes; performance in other environments has no examples.
- The image will include the Renault logo and RENAULT / GORDINI text. Commercial use requires attention to trademarks. Additionally, 2 sensitive-level example images have been hidden.
Trigger Word: Renault 8 Gordini Concept
[A] Casual Candid & Ultra Realistic Phone Snap · MOSS
LoRA · Ported · Author 858654128784 · Published 11-19 18:12 · Downloads 92
"Candid phone snapshot" style LoRA: Subjects do not look at the camera, composition is tilted, foreground has obstructions, lighting is uneven, and details are rough. There are specific patterns for near / medium / far shot distances.
Images

Reasons for Recommendation
- Ported: The author claims that the MOSS and Krea 2 versions were released simultaneously using the same dataset. The entire model page has 628 downloads. Differences between the two versions are not specified by the author.
- Serves "non-posed" realistic photos for daily scenarios such as dorms, commuting, and offices, removing the posed feel from AI images.
- The author provides a complete prompt order: Person + Action + Location → Shoes/Socks → Background → One foreground obstruction → Tilted composition, rough digital details, uneven lighting. 9 visible examples. 92 downloads in 3 hours since release.
How to Use
- No trigger word. Write according to the author's prompt order, ending with "tilted framing, coarse digital detail, uneven lighting".
- Specify what the person is looking at (phone, book, shoelaces); do not have them look at the camera. Describe the camera position reasonably (door crack, across the table, behind, top bunk); do not describe a phone held above the face.
- LoRA strength, steps, and CFG are not provided by the author.
Risks and Limitations
- Visible examples: 9 images, with another 11 sensitive-level images hidden; the dataset is predominantly male (the description consistently uses "he")
- The theme mimics a candid photography perspective; using it for real people or passing off generated images as real candid photos carries ethical and legal risks
[A] Moody Moss Mix (Free download 免费下载) · V5.5 15STEPS CFG3
Base Model / Checkpoint · Version Update · Author catlover1937 · Published 11-19 17:20 · Downloads 350
The V5.5 free quantized version of the photorealistic MOSS fusion base model Moody Moss Mix (available in FP8 and INT8 files), featuring a built-in new acceleration LoRA, retuned for image generation at approximately 15 steps.
This is the same V5.5 release as the paid full-precision version on the same page, just in different precisions; choose this free version for daily image generation. The author stated in the previous version's notes that the visual difference between the quantized and full-precision versions is imperceptible to the naked eye.
Images

Reasons for Recommendation
- Version update (per author): Multiple underlying LoRAs were redone, resulting in more natural body structures and material textures; switched to a new Turbo LoRA with weights tuned for image generation at approximately 15 steps; further reduced bias toward Asian ethnicities; reduced skin oiliness; overall exposure is slightly darker than the previous generation, closer to photography
- Serves photorealistic portraits and photographic-style text-to-image generation; the author claims that when combined with the 4K HD workflow updated on the same day, overall generation speed is approximately 35% faster than the previous generation
- Free download available on this page: FP8 and INT8 are each approximately 4.7 GB; the full-precision version of the same V5.5 is another paid entry on the same page; published 4 hours ago with 350 downloads (indicating popularity only)
How to Use
- Replace the MOSS base model for loading; choose either FP8 or INT8; use 15 steps and CFG 3 as per the version name
- The author did not specify a sampler for V5.5 (the old version V3 recommended Euler Simple); no trigger word is required
- The author recommends using it together with the same author's Moody Moss 4K HD Workflow V4
Risks and Limitations
- Only 2 visible examples, with another 18 sensitive-level images hidden; there are no visible before-and-after comparisons for the changes in the new version
- The author describes this as an experimental major overhaul and welcomes feedback; the version notes section is empty, with changes listed at the top of the model card
- The license does not allow derivatives, only on-site rental; most of the model card consists of promotional content such as sponsorships and invitation codes
- See 3479331 for the BF16 full-precision paid version of the same V5.5
[A] Dizzle's - The Sun · v1
LoRA · New Model · Author Maridizzle · Published 11-19 12:35 · Downloads 29
Places a glowing sun in the sky of the image, tints the overall scene with warm orange light, and adds a burnt paper edge texture; the second entry in the author's "Outer Space" series.
Images

Reasons for Recommendation
- New model: The author claims it is suitable for portraits, still life, witches, interiors, and even a cactus, producing an effect of warm orange lighting with burnt paper edges
- Serves warm-toned illustrations, retro paper-textured posters and cards; effective just by writing the trigger word
- Trigger word and strength are provided; 5 visible examples; permissive license (generation, rental, derivatives allowed, no attribution required)
How to Use
- Write dz_sun at the beginning of the prompt; strength 0.9-1.0; no additional keywords needed
- Steps, CFG, and sampler were not provided by the author
Risks and Limitations
- The image will be uniformly tinted orange with paper edges, making it unsuitable for scenes requiring original color preservation; 6 sensitive-level examples are hidden
Trigger Word: dz_sun
[A] Dizzle's Venus · v1
LoRA · New Model · Author Maridizzle · Published 11-19 14:49 · Downloads 25
A "granularization" concept LoRA that transforms the subject, clothing, and background into flowing ribbons composed of tiny glittering beads (orange, lavender, cream stripes), accompanied by floating bubbles and a glowing teal background.
Images

Reasons for Recommendation
- New model: The author claims it is suitable for portraits, interiors, landscapes, animals, and objects, transforming the entire image into beaded ribbons
- Serves decorative illustrations and dreamy-style posters; the effect's color palette is fixed, representing strong stylization
- Trigger word and strength are provided; 4 visible examples; permissive license (generation, rental, derivatives allowed, no attribution required)
How to Use
- Write the trigger word at the beginning of the prompt, with a capital V as requested by the author: dz_Venus; strength 0.9-1.0
- Steps, CFG, and sampler were not provided by the author
Risks and Limitations
- Inconsistent capitalization in the trigger word: The description says dz_Venus, while the site registration is Dz_venus; following the description is safer
- The color palette is locked to orange, lavender, cream, and teal; 3 sensitive-level examples are hidden
Trigger Word: Dz_venus
[A] Dizzle's Drip Storm · v1
LoRA · New Model · Author Maridizzle · Published 11-19 11:45 · Downloads 33
A "rainstorm drip" concept LoRA that adds fine dark threads and sparkling droplets to any image, accompanied by wet reflective surfaces and heavy dim lighting.
Images

Reasons for Recommendation
- New model: The author claims this dripping thread and droplet effect can be added to characters, creatures, and even entire cosmic scenes
- Serves dark, atmospheric posters and concept art; only the trigger word is needed, no additional prompts required
- Trigger word and strength are provided; 5 visible examples; permissive license (generation, rental, derivatives allowed, no attribution required)
How to Use
- Write dz_drip at the beginning of the prompt, followed by the subject; strength 0.9-1.0
- Steps, CFG, and sampler were not provided by the author
Risks and Limitations
- The description is only one paragraph, with no failure cases or behavior when strength is too high; 1 sensitive-level examples are hidden
Trigger Word: dz_drip
[A] Tintype Portrait · MOSS
LoRA · New Model · Author 858654128784 · Published 11-19 13:32 · Downloads 64
A wet plate collodion (tintype) texture LoRA: dark plate frames, deep brown shadows and silver highlights, white deposits on the plate edges, transforming modern people and scenes into century-old photographs.
Images

Reasons for Recommendation
- New model: One of several concept LoRAs by the same author in this period; the author claims farms, barns, platforms, and modern people can all be transformed into century-old photo textures, with focus on the eyes
- Serves retro portraits, historical theme posters, and character/scene images in old photo style; the author claims half-body close-ups are the most realistic
- Trigger word, prompt sentence structure, and strength are provided; 5 visible examples, published 8 hours ago with 64 downloads, free, permissive license
How to Use
- Start the prompt with tintypeportrait, followed by One figure, ... adult man ..., faces the camera in ... (the trigger word is not registered in the version and must be written manually)
- Steps, CFG, and sampler were not provided by the author
Risks and Limitations
- Only 5 visible examples, with another 15 sensitive-level images hidden
[A] Billboards · MOSS
LoRA · Ported · Author 858654128784 · Published 11-19 11:54 · Downloads 77
A city outdoor large billboard LoRA: rooftop steel frame signs, roadside single-pole signs, commercial street LED screens, from a casual smartphone photography perspective, with small pedestrians and traffic below to indicate scale.
Images

Reasons for Recommendation
- Ported: A Krea 2 version already exists on the same page, with 379 downloads for the entire model page; the MOSS version has no version notes, and differences from the original are not specified by the author
- Serves advertising mockups, city street scenes, and brand poster scenarios; people, slogans, and brands in the ads are specified in the prompt, and blank signs or signs with changed images can also be generated
- The author provided text tips: Uppercase short words of 5–7 letters work best; 4 visible examples, published 9 hours ago with 77 downloads, free, permissive license
How to Use
- Trigger word billboardconcept (written in the description, not registered in the version, must be written manually)
- Write ad characters, slogans, and brands into the prompt, using uppercase words of 5–7 letters for slogans
- LoRA strength, steps, and CFG were not provided by the author
Risks and Limitations
- The description is only one paragraph, with no parameters; performance with long sentences or lowercase text is not specified
- 4 visible examples, with another 4 sensitive-level images hidden
[A] Yuutenji_Nyamu_MOSSimage_lora · v1.0
LoRA · New Model · Author yueyinganshuang · Published 11-19 20:50 · Downloads 4
A character LoRA for Yuutenji Nyamu, covering two outfits: a gray off-shoulder casual outfit and a black-red tailcoat stage outfit; the author provided complete prompts that can be copied directly.
The most complete documentation in this group; simply copy the base model, encoder, VAE, sampling, and resolution from the card to reproduce the results.
Images

Reasons for Recommendation
- New model: The author claims it was trained on 300 images, as the base model itself does not recognize this character; complete prompts describing hairstyle, eye color, and full clothing are provided for each of the two outfits, ready to paste and use
- Serves fan art illustrations and character standing pose text-to-image generation; the reproduction environment is fully documented: SFT int8 base model, qwen3.5 text encoder, flux2-vae, sampling parameters, and full-body/half-body resolutions, along with known issues (the number of buttons on the casual outfit primarily depends on the seed)
- 4 visible examples, free download; published less than 1 hours ago, so download counts are not yet meaningful
How to Use
- The prompt must start with Yuutenji Nyamu, followed by pasting the description for either the casual or stage outfit; actions, shot types, and backgrounds are appended after commas
- Author parameters: 30 steps, CFG 4.0, euler + sgm_uniform, denoise 1.0, LoRA strength 1.0
- Resolutions: Full body 1216×1616, half body 1216×1216; base model moss_image_sft_int8_convrot, VAE uses flux2-vae
- Negative prompt (provided by author): blurry, low quality, deformed, wrong text, garbled characters, watermark
Risks and Limitations
- Standard copyrighted character LoRA, covering only two outfits; only 4 examples
- Trigger word not registered (the page does not display a copyable trigger word), so the character name must be written manually as per instructions; license does not allow derivatives
- Author note: If ComfyUI output is poor, upgrade to the latest version
Worth Watching (13)
[B] SidelightGAL: Mya — Realistic Aisan Girl | MOSS-Image Character LoRA · v1.0
LoRA · New Model · Author yinhades · Published 11-19 11:20 · Downloads 75
A character consistency LoRA for a realistic fictional Asian female: one trigger word locks the same face and body type, while hairstyle, clothing, pose, and scene change according to the prompt.
Images

[B] Halsin - BG3 · Halsin_Moss
LoRA · Ported · Author AthosAI · Published 11-19 16:01 · Downloads 14
A character LoRA for Halsin, the Archdruid from Baldur's Gate 3 (long brown braided hair, pointed ears, red facial markings), including part-by-part prompts for his signature druid armor.
Images

[B] Hyper Realistic Anime · Moss 1.0
LoRA · Ported · Author JBr4t · Published 11-19 12:40 · Downloads 12 · Paid
A LoRA for rendering anime characters in a “realistic anime” style with near-photorealistic texture; the MOSS version is ported from the Illustrious version and retrained.
Image

【B】Dizzle's Planet Earth · v1
LoRA · New Model · Author Maridizzle · Published 11-19 15:47 · Downloads 24
Scene LoRA: Uses fixed phrases to place the subject in one of five Earth scenes—polar bays, alpine lakes, tropical rainforest rivers, islands, or dusk port cities.
Image

【B】Dizzle's Black Hole Effect · v1
LoRA · New Model · Author Maridizzle · Published 11-19 12:29 · Downloads 33
Places the subject in a black void scattered with stars; this is the first entry in the author’s “Outer Space” series. Indoor scenes can also be transformed into starry ceilings.
Image

【B】Lost Paradisa [MOSS] · v1.0
LoRA · New Model · Author Franklin_Sports · Published 11-19 15:25 · Downloads 56
Pictorialism-style LoRA: Draws inspiration from British and American photographers’ works from the 1960–70s, featuring soft focus and dreamlike lighting for an old-photo aesthetic.
Image

【B】Dizzle's Mercury · v1
LoRA · New Model · Author Maridizzle · Published 11-19 13:10 · Downloads 24
A “Mercury” style LoRA that renders images as thick palette-knife oil paintings, dominated by reds and oranges with cool teal-green shadows.
Image

【B】Silvana & Tamra's Style Mix · Moss V.1
LoRA · Ported · Author TamraLesgarden · Published 11-19 18:22 · Downloads 20
A “secondary distillation” version of the two authors’ collaborative style: Retrained on images generated by their Style Mix LoRA, aiming for shorter prompts.
Image

【B】AuthentiQ 3D: Pixar Style [Krea2/MOSS] · MOSS_v1.0
LoRA · Ported · Author AuthentiQ · Published 11-19 11:30 · Downloads 73
A Pixar-style 3D animated character LoRA, ported from the Krea 2 version to MOSS SFT.
Image

【B】Silver Rift · MOSS-Image-base
LoRA · New Model · Author AIImageStudio · Published 11-19 20:18 · Downloads 1 · Paid
A photographic style LoRA with slide film (positive film) tones: stronger contrast and saturation, with more dramatic light and shadow.
Image

【B】Frutiger Aero · v1.0
LoRA · New Model · Author eichenwaldatelier461 · Published 11-19 12:03 · Downloads 27
Frutiger Aero art style LoRA (based on the name and tags, this refers to the glossy glass, bubbles, blue-green nature, and tech aesthetic of the Windows Vista / 7 era); the author only noted “same dataset.”
Image

【B】Bogs and Subarctic Terrains · MOSS SFT
LoRA · Ported · Author quintronanthony265 · Published 11-19 16:17 · Downloads 16
Northern subarctic wetland terrain LoRA: Adds northern ecological backgrounds such as peat bogs, tundra, taiga forests, and shrublands to the image.
Image

【B】60s Psychedelic Movie · Moss
LoRA · Ported · Author ArsMachina · Published 11-19 19:10 · Downloads 1 · Paid
A 60s psychedelic movie still style LoRA, a MOSS port of the model page from an established multi-base-model provider, offering both Base and SFT weights.
Image

Included Only (24)
- yul-016-(Hinoa) moss-Monster Hunter · yul-016-(Hinoa) moss — LoRA: A character LoRA for Hinoa, the receptionist from Monster Hunter Rise (black hair, pointed ears, red and gold Japanese-style attire), with the author noting a BJD doll texture.

- Hiyori / Japanese AI girl · MOSS — LoRA: A character LoRA for Hiyori, a fictional Japanese woman (described by the author as an AI-generated character), using the trigger word kz_hiyori to maintain consistency.

- Arezu - ヒナツ - Pokemon - MOS LORA · v1.0 — LoRA: A character LoRA for Arezu from Pokémon (Japanese name Hinatsu: short reddish-orange spiky hair, red eyes), including a prompt set for a dark blue hoodie outfit.

- Kallen Kouzuki - Code Geass - MOSS LORA · v1.0 — LoRA: A character LoRA for Kallen Kouzuki from Code Geass (red hair, blue eyes), registering prompts for two outfits: a red and grey pilot suit and casual wear with a red headband.

- Moss Image Amber — Fictional Character · v1 — LoRA: A character LoRA for Amber, an original fictional adult woman, using the trigger word amber-notarealperson to maintain identity consistency.

- Yoruichi Shihouin - Bleach: Thousand-Year Blood War - MOSS LORA · v1.0 — LoRA: A character LoRA for Yoruichi Shihouin from Bleach: Thousand-Year Blood War, registering four prompt segments: the character, a black bodysuit, an orange skirt with a white cloak, and her "Shunpo" lightning form.

- Silverwolf HSR · v1.0 — LoRA: A character LoRA for Silver Wolf from Honkai: Star Rail; the author states that MOSS can already draw her, so this LoRA is only for detail refinement.

- Moody Moss Mix (Free download 免费下载) · V5.5 15STEPS CFG3 BF16 — Base Model / Checkpoint: The full-precision BF16 version of Moody Moss Mix V5.5 (approx. 9.5 GB, paid), with weights identical to the free FP8 / INT8 versions released on the same day.

- Lower lip bitting MOSS · v1 — LoRA: A single-action LoRA that makes characters perform a "lower lip biting" expression.

- xTrinia · MOSS Image — LoRA: A character LoRA for xTrinia, a fictional mature African-American woman; the author claims all data is AI-generated and no real photos were used.

- Darcey Von Vergin · MossImage — LoRA: A character LoRA for Darcey Von Vergin, the fictional female president of a fashion company; the MOSS version is another entry in the multi-base-model series on the same page.

- OpenAI Chatgpt Anime Style - MOSS · v1 — LoRA: A style LoRA described by the author as a tribute to the ChatGPT anime art style; the model card contains only one sentence.

- Pearl (Honkai: Star Rail) · v1.0 Moss — LoRA: A character LoRA for Pearl from Honkai: Star Rail; the MOSS version is ported from the Anima and Krea 2 versions on the same page.

- Light and Shadow Floating Dream Semi-Realistic Style Lora · v1.0 — LoRA: A semi-realistic art style LoRA named "Light and Shadow Floating Dream"; the author's description is only one sentence, so the specific style must be judged from the examples.

- Matcha Specialty Shop Matcha Style Lora · v1.0 — LoRA: An art style LoRA named "Matcha Specialty Shop Matcha"; the author only wrote "style lora, not for commercial use," so the style must be judged from the examples.

- arisu_terui (seitokai ni mo ana ha aru) · v1.0 (moss) — LoRA: A character LoRA for Arisu Terui from the manga Seitokai ni mo Ana wa Aru! (There's a Hole in the Student Council Too!), ported from the Anima version to MOSS.

- Power Girl (DC Comics) [MOSS Image] · v1.0 — LoRA: A character LoRA for Power Girl from DC Comics (short curly blonde hair, white bodysuit, red cape, blue gloves and boots).
- Emilia purple nightgown | Re:Zero | MOSS-Image LoRa · Purple nightgown — LoRA: An outfit LoRA for Emilia's purple nightgown from Re:Zero - Starting Life in Another World, with the trigger word eminightgown.

- MOSS GPT · v1.0 — LoRA: A single-character LoRA named "gpt" (black high-neck short vest, silver-white coat, black pleated mini-skirt, black floral hair accessory); the author did not specify the character's origin or purpose.

- mikoto (itoko_no_ko) · v1.0 (moss) — LoRA: A character LoRA for Mikoto Yamato from the manga Itoko no Ko, ported from the Anima version to MOSS.

- Octane Render · MOSS — LoRA: The MOSS port of the classic Octane Render style LoRA (paid): cinematic ultra-detail, glossy surfaces, close-up portraits; this version includes separate SFT and Base files.

- Sea Phantom Politis (Epic Seven) · Moss Image — LoRA: A LoRA for Sea Phantom Politis from Epic Seven, registering prompts for two outfits: a gothic long dress and a ballet tutu, ported from the Anima / Illustrious / Pony versions.

- Velina Airgid ヴェリナ・エガード, 2 Outfits — Zenless Zone Zero ゼンレスゾーンゼロ [Anima+Illustrious+MOSS] · v1.0 MOSS Image — LoRA: A LoRA for Velina Airgid from Zenless Zone Zero, with one file containing two outfits (default and Shade of Leisure), ported from the Anima / Illustrious versions.

- HMA-anime style · v1.0 — LoRA: An anime art style LoRA named HMA; the model card contains only the trigger word hma_anime, with no description of the style.

Data: Civitai public API (base model MOSS Image), collection window 11-19 08:31 → 11-19 21:01 (Beijing Time). Recommendations have been reviewed by editors; download counts only reflect popularity.