MOSS Image Evening Report 2026-11-18 | Top Pick: MOSS Flat-illustration style

This issue reviewed 21 new MOSS versions: 1 S, 4 A, 4 B, 11 C, with 1 excluded. Top pick: Flat-illustration style LoRA, which compresses character shading and volume into clean color blocks while keeping backgrounds photorealistic. The author provided a weight tier table from 0.5 to 2.0; both visible examples are multi-weight comparison images using the same prompt. Serpieri European comic pen-and-watercolor style, Gulfstream G600 private jet interior, shrunken people scale contrast concept, and Millennium flash snapshot / Y2K product photo style are rated A; JIKO anime INT8 base model, Blue Archive style MOSS new version, Little Nicholas style, and high-heel slip-off action LoRA are rated B; standard character/virtual person LoRAs, paid LoRAs with empty descriptions, and those with only examples and unclear purpose are rated

Zhenzhen's AI Radar · MOSS Image · 2026-11-18 Evening Report

This issue features 18 new versions: Strongly Recommended 1 · Recommended 3 · Worth Watching 3 · Included Only 11

Editor's Note This issue reviewed 21 new MOSS versions: 1 S, 4 A, 4 B, 11 C, with 1 excluded. Top pick: Flat-illustration style LoRA, which compresses character shading and volume into clean color blocks while keeping backgrounds photorealistic. The author provided a weight tier table from 0.5 to 2.0; both visible examples are multi-weight comparison images using the same prompt. Serpieri European comic pen-and-watercolor style, Gulfstream G600 private jet interior, shrunken people scale contrast concept, and Millennium flash snapshot / Y2K product photo style are rated A; JIKO anime INT8 base model, Blue Archive style MOSS new version, Little Nicholas style, and high-heel slip-off action LoRA are rated B; standard character/virtual person LoRAs, paid LoRAs with empty descriptions, and those with only examples and unclear purpose are rated C. 'MOSS Simple Comfy Workflow (repost)' is a repost of the author's most basic text-to-image workflow JSON, not a new weight, so it is excluded.

Strongly Recommended (1)

[S] MOSS Flat-illustration style · v1.0

LoRA · New Model · Author baowy · Published 11-18 14:05 · Downloads 58

Flat-illustration style LoRA: Removes character shading and volume, turning characters into clean color blocks, moving towards webtoon/vector illustration styles, while backgrounds remain photorealistic.

Suitable for directly replicating the author's four-tier comparison: Run the same prompt at 1.0 / 1.5 / 2.0 to see if the background remains photorealistic while the character becomes flat.

Images

Example 3136x4326

Reasons for Recommendation

  • New model; the author has documented the effects in a weight tier table: 0.5 remains semi-realistic with soft shading, 1.0 is the balance point, 1.5 is where the flat feel truly emerges, and 2.0 is the flattest but with noticeable detail loss—users can directly select the intensity based on the desired level of flatness
  • All 2 visible examples are four-tier weight comparison collages using the same prompt and seed (one close-up portrait, one full-body street crouch), showing that the background street scene consistently maintains photorealistic quality while only the character becomes flat, consistent with the author's description
  • Serves posters, webtoon covers, and character promotional images featuring 'flat anime characters standing in photorealistic scenes'; follows the same approach as the author's previous cel-shading style LoRA (changing only the character, not the background), but this time using flat color blocks
  • Free, license allows image generation, selling generated images, and merging; no trigger word required; published approximately 7 hours ago with 58 downloads (download count only indicates attention level)

How to Use

  • Load the LoRA, no trigger word required (not registered by the author); recommended weight is 1.0, increase to 1.5 for a distinct flat feel, with 2.0 as the upper limit
  • The author states that pairing with a 8-step DMD acceleration LoRA (strength 1.0) yields the best results; sampler and CFG were not provided by the author

Risks and Limitations

  • Author's warning: At weights above 1.5, mechanical and armor details will collapse into color blocks; weights between 0.5 and 1.0 still look like semi-realistic illustrations, so do not assume they are ineffective
  • Only 2 comparison collages are visible (weights labeled as 0.7 / 1 / 1.5 / 2); another 2 sensitive-level examples are hidden; the subject matter is limited to young female portraits, and performance on other subjects is unknown

Recommended (3)

[A] Gulfstream G600 Interior [Moss] · v1

LoRA · New Model · Author denrakeiw · Published 11-18 12:20 · Downloads 41

Gulfstream G600 private jet (N600G) interior and cockpit LoRA: Trained on official interior photos, capable of generating cream-colored herringbone leather seats, high-gloss dark oak surfaces, oval windows, mid-cabin dining tables, aft-cabin sofas, and the Symmetry glass cockpit.

Images

Example 1776x2368

Reasons for Recommendation

  • New model; serves private jet interior images, luxury travel advertisements, and interior design references; the author has written the objects in each cabin zone (seat materials, wood veneers, window shades, tableware, pillows, cockpit screen layout) into directly pasteable prompt fragments
  • The complete model card also provides five example prompts written in the training set's style (cabin aisle, mid-cabin dining table, aft-cabin lounge, sofa close-up, cockpit), which can be copied directly to reproduce the corresponding compositions
  • Among the 10 visible examples, besides standard cabin photos, there are images turning the cabin into posters, sales advertisements, and snack ranking layouts, indicating compatibility with text-layout prompts
  • Free, license allows image generation, selling generated images, and merging; published approximately 9 hours ago with 41 downloads (download count only indicates attention level)

How to Use

  • Trigger phrase: Gulfstream G600 business jet; the author suggests starting prompts with 'Create a photograph of ... inside the Gulfstream G600 business jet'
  • Strength 0.85–1.0; then append the author's object terms based on the cabin zone (e.g., main passenger cabin, aft stateroom lounge, Gulfstream Symmetry Flight Deck cockpit)
  • Steps, CFG, and sampler were not provided by the author

Risks and Limitations

  • Covers only one specific three-cabin layout for this aircraft, resulting in a narrow use case; training images are official Gulfstream photos, and generated images involve real brands and models, so pay attention to trademark issues before commercial use
  • The page contains mixed content, and 10 sensitive-level examples are hidden; the model card in data.json is truncated, and the prompt opening sentences and example prompts are taken from the complete model card

Trigger Word: Gulfstream G600 business jet

[A] Shrunk People · MOSS

LoRA · Ported · Author 858654128784 · Published 11-18 10:02 · Downloads 100

MOSS version of the shrunken people concept LoRA: Shrinks people to a few centimeters tall, standing next to desks, shoes, or palms, turning everyday objects into giants to create photorealistic images with scale contrast.

Images

Example 832x1216

Reasons for Recommendation

  • Ported: The same model previously had a Krea 2 version; the MOSS version uses the same description without noting differences from the Krea 2 version
  • Serves fun creative images, advertisements, and social media graphics; all 10 visible examples are photorealistic, showing tiny people standing next to sneakers, socks, coffee cups, palms, or toothpaste, or emerging from backpacks, making the concept immediately clear
  • Free, license allows image generation, selling generated images, and merging; published approximately 11 hours ago with 100 downloads, and the entire model has 603 downloads (download count only indicates attention level)

How to Use

  • Load the LoRA, no trigger word registered by the author; write scenes using the sentence structures in the description, such as 'a tiny man on a desk / beside a shoe / on a palm', and specify the adjacent everyday objects
  • Strength, steps, and CFG were not provided by the author

Risks and Limitations

  • The description is only one sentence with no parameters; the filename is shrunkmaleconcept, and all examples feature male tiny people, so the effect for female or multiple characters is not specified

[A] Millennium · MOSS

LoRA · Ported · Author 858654128784 · Published 11-18 19:08 · Downloads 29

MOSS version of the Millennium era (early 2000s) photo style LoRA, featuring two vibes: amateur lifestyle snapshots with direct flash and fine grain (floral bed sheets, CRT TVs, internet cafes, KTVs, cassettes, and CDs), and glossy Y2K product photos of translucent teal/purple computers, phones, and MP3s.

Images

Example 896x1152

Reasons for Recommendation

  • Ported: The same model previously had a Krea 2 version; the MOSS version uses the same usage instructions without noting differences from the Krea 2 version
  • Usage instructions clearly define the sentence structure: Start with 'millenniumera' and write a complete sentence, ending with a lighting + style sentence to switch between the two vibes; the author claims it works for people, groups, still life, and empty rooms
  • The 3 visible examples show two people in front of internet cafe computers, a portrait of a person holding an MP3 next to a translucent computer, and a person next to a CRT TV in an old-style living room, matching the two described styles
  • Free, license allows image generation, selling generated images, and merging

How to Use

  • Start the prompt with 'millenniumera' and describe the scene in a complete sentence
  • For lifestyle snapshots, end with 'Direct flash and fine grain create an early-2000s snapshot.'; for product photos, end with '... create a glossy millennium-era photograph.'
  • Strength, steps, and CFG were not provided by the author

Risks and Limitations

  • The trigger word 'millenniumera' is only mentioned in the description and not registered in the trigger word field
  • Only 3 visible examples are available, with another 5 sensitive-level examples hidden; published approximately 2 hours ago, so feedback is still limited

Worth Watching (3)

[B] Blue Archive Style (blue archive style) style 【anima/ill/krea2】 · moss heal

LoRA · Version Update · Author umina · Published 11-18 14:37 · Downloads 67

Second MOSS version (moss heal) of the Blue Archive official art style LoRA, trained on over 400 BA official images to produce flat-shaded anime characters in the style of BA character art.

Images

Example 1024x1536

Reasons for Recommendation

  • Version update: The previous MOSS version registered the trigger word '@ba nai'; this version's filename changed to v2, no trigger word is registered, the version description is empty, and it is unclear what changed or which base model 'heal' refers to
  • Serves BA-style character art and fan art; most of the 11 visible examples are flat-shaded girl portraits on white or simple backgrounds, with two featuring black-and-white characters on colored backgrounds, though the author does not explain if this is intentional
  • The entire model page has 11495 downloads and 619 likes, mostly from other base model versions like Anima, Illustrious, and Krea 2 (download count only indicates attention level)

How to Use

  • After loading the LoRA, write anime character prompts; the auxiliary tags registered by the author are 'blue archive, three quarter view, dark, red theme, warm colored, cold colored'
  • Strength, steps, and CFG for the MOSS version were not provided by the author; the author suggested slightly lowering the weight to 0.7-0.8 for the Anima version, which can serve as a starting reference

Risks and Limitations

  • The author did not write an update description; it is unclear if the previous version's '@ba nai' trigger word is still required
  • Trained on copyrighted game official art; the filename contains 'nai', and the author does not explain if some training data comes from NovelAI-generated images
  • The license does not allow derivatives, and commercial use is only checked for on-site rental; 4 sensitive-level examples are hidden

[B] Le Petit Spirou - Janry - Style - MOSS LORA · v1.0

LoRA · New Model · Author Konan · Published 11-18 10:59 · Downloads 60

MOSS LoRA for the style of the Belgian comic 'Le Petit Spirou' (artist Janry): Franco-Belgian comic-style cartoon characters with thick outlines and flat color blocks.

Images

Example 1280x1840

Reasons for Recommendation

  • New model; serves Franco-Belgian comic-style character illustrations and comic covers; same author previously released LoRAs for styles like SamDoesArts and MoriiMee, using the same syntax
  • The only visible example is a blonde girl resting her chin on her hand against a yellow background; the thick ink lines and flat coloring indicate a Franco-Belgian comic direction
  • Trigger phrase and strength are provided; free, license allows image generation; released approximately 10 hours ago with 60 downloads and 10 likes (download count only indicates interest)

How to Use

  • Prompts start with "Illustration of" (the registered trigger word is "illustration of"); strength 0.8-1.3
  • Steps, CFG, and sampler are not specified by the author

Risks and Limitations

  • Only 1 visible examples, with 6 sensitive-level images hidden; the model card only lists the training base model, strength, and trigger phrase
  • Mimics copyrighted comics and real artists' styles, with no mention of authorization by the author; the trigger word "illustration of" is very generic, making it indistinguishable when used alongside other style LoRAs with the same trigger phrase

Trigger Word: illustration of

[B] Moss LORA ShoePlay-Heelpop · v1.0

Base Model / Checkpoint · New Model · Author pandemicmonkee · Released 11-18 12:48 · Downloads 37

Heelpop LoRA for high heels: depicts the action of a woman's heel slipping out of the shoe, with the shoe hanging on the toes; covers various shoe types, works with stockings, nylon socks, or bare legs; the page type is labeled as a base model, but the file is actually a LoRA.

Images

Example 1024x1536

Reasons for Recommendation

  • New model; serves this specific action in footwear/leg close-ups; the author states that changing the prompt can also produce the look of shoes dangling on the toes (dipping)
  • Trigger word, starting strength, and training steps are all provided; the author explains it was trained on Moss Base and is usable on Moss SFT
  • The only visible example is a close-up of a heel slipping out of a black stiletto under an office desk, consistent with the description; free

How to Use

  • Load as a LoRA (file size approx. 138 MB, same size as other MOSS LoRAs this period, cannot be used as a base model); write "heelpop" in the prompt
  • Adjust strength from 0.4 upwards to above 1.0 until the heelpop amplitude is appropriate; the "Steps: 2000" written by the author refers to training steps, while generation steps and CFG are not provided by the author

Risks and Limitations

  • Very narrow use case, with only 1 visible examples; the trigger word "heelpop" is only mentioned in the description and not registered in the trigger word field
  • This period's data recorded the type as Checkpoint, but the current version page has been changed to LoRA; the license only checks for on-site rental and rental, not allowing commercial use of generated images

Included Only (11)

  • Chiaki / Japanese AI girl · MOSS — LoRA: MOSS version of a realistic character LoRA for Chiaki, a fictional Japanese female (AI-generated character), trained on MOSS-Image-Base, intended for single-person images. Example 1024x1536
  • Qtstyle · v0.01 — LoRA: v0.01 beta of the author's first LoRA: aims for 2.5D stylized character images with "proportion control"; the only visible example is an elf girl in a gothic classroom. Example 1344x1760
  • Sheena Fujibayashi (Tales of Symphonia) | 3 Outfits [MOSS Image] · v1.0 — LoRA: MOSS character LoRA for Sheena Fujibayashi from "Tales of Symphonia", including three sets: default outfit, beach swimsuit, and purple dancer outfit. Illustration 1672x941
  • Lady Dimitrescu · MOSS Image · Resident Evil Village · v1.1 — LoRA: MOSS character LoRA v1.1 for Lady Dimitrescu (Alcina Dimitrescu) from "Resident Evil Village"; update notes state that training steps were added and a larger, cleaner dataset was used. Illustration 1672x941
  • Moss LORA Candies Mules · v1.0 — Base Model / Checkpoint: Apparel LoRA for retro Candie's high-heeled mule slippers: leather or suede straps, various colors, can also produce transparent PVC straps; the page type is labeled as a base model, but the file is actually a LoRA. Illustration 1254x1254
  • kosaki onodera · v1.0 (moss) — LoRA: MOSS character LoRA for Kosaki Onodera from "Nisekoi"; the author notes that the MOSS base model itself cannot draw this character. Example 498x728
  • Class37_lady_victoria · v1 — LoRA: Realistic female character LoRA, with the description only stating "Female model"; the 2 examples are lifestyle photos of a blonde woman. Example 512x512
  • SoulTide 灵魂潮汐 Philodoxy 菲洛多希 DiT · MOSSTagFlowV1.1 — LoRA: MOSS port of the character LoRA for Philodoxy (pink-haired, purple-eyed elf) from the mobile game "Soul Tide", including four sets of prompts: casual wear, Jail, idol outfit, and pure appearance. Illustration 1254x1254
  • Chisy (AI Model) [Minimax / MOSS / Krea2 / SD] · MOSS — LoRA: MOSS version of a realistic character LoRA for the virtual AI influencer Chisy: wolf cut hair with split dyeing (one side orange to light gold, the other dark brown with wine red), gray-blue eyes, and thick black eyeliner. Example 896x1184
  • CYBERCORE BY STX (Moss) · X1 — LoRA: Paid cyberpunk "assistant" concept LoRA; the author only states it is used to generate high-quality cyberpunk images, with a strength of approximately 1.00. Example 768x1344
  • 丹丨蛊 — Twin Poles [Style Mix No.01] · v1.0 — LoRA: The first "Twin Poles" in the "丹丨蛊" style mix LoRA series: the author fused multiple style LoRAs and selected the first 12 stable combinations from thousands of tests; this is one of them. Example 2240x1408

Data: Civitai Public API (base model MOSS Image), inclusion window 11-18 08:31 → 11-18 21:01 (Beijing Time). Recommendation reasons have been editorially reviewed; download counts only indicate interest.

Civitai

Explore the latest AI Radar updates