MOSS Image Evening Report 2026-11-16 | Top Pick: Nix Comic [Multiple Styles]

This issue reviews 22 new MOSS versions, including 1 S-tier, 8 A-tier, 8 B-tier, and 5 C-tier entries. The top pick is the MOSS version of Nix Comic (Disney / Comic): a comic-style LoRA based on 90s Disney animation and 00s American comics foundation. No trigger word is required; both strength settings and step counts for accelerated/standard workflows are provided, along with solutions for when character LoRAs wash out the style. The same author released an MOSS-only Euro comic variant on the same day, listed as A-tier. SFT INT8/INT4 mixed quantization runnable on 6 GB GPUs, Layout Studio v1.1.0 which generates bbox JSON prompts via drag-and-drop boxes, and a concept LoRA for Chinese summer snapshot style from around the year 2008 are listed as A-tier; two items labeled Other—pose control and Canny style modification workflows—were verified to be MOSS content and are listed as B-tier workflows; two paid STX concept LoRAs and an artist-style LoRA with only a one-sentence description are listed as C-tier.

Zhenzhen's AI Radar · MOSS Image · 2026-11-16 Evening Report

This issue features 19 new versions: Strongly Recommended 1 · Recommended 8 · Worth Watching 7 · Listed Only 3

Editor's Note This issue reviews 22 new MOSS versions, including 1 S-tier, 8 A-tier, 8 B-tier, and 5 C-tier entries. The top pick is the MOSS version of Nix Comic (Disney / Comic): a comic-style LoRA based on 90s Disney animation and 00s American comics foundation. No trigger word is required; both strength settings and step counts for accelerated/standard workflows are provided, along with solutions for when character LoRAs wash out the style. The same author released an MOSS-only Euro comic variant on the same day, listed as A-tier. SFT INT8/INT4 mixed quantization runnable on 6 GB GPUs, Layout Studio v1.1.0 which generates bbox JSON prompts via drag-and-drop boxes, and a concept LoRA for Chinese summer snapshot style from around the year 2008 are listed as A-tier; two items labeled Other—pose control and Canny style modification workflows—were verified to be MOSS content and are listed as B-tier workflows; two paid STX concept LoRAs and an artist-style LoRA with only a one-sentence description are listed as C-tier.

Strongly Recommended (1)

[S] Nix Comic [Multiple Styles] · Disney / Comic [MOSS]

LoRA · Port · Author xFennec777 · Published 11-16 18:06 · Downloads 42

The MOSS version of Nix Comic (Disney / Comic): An American comic-style LoRA based on 1990s Disney animation and 2000s Marvel comics. The author specifically documented step counts and notes for stacking LoRAs on MOSS.

Images

Example 1248x1680

Reasons for Recommendation

  • Port: The page previously featured multiple Krea 2 versions; the MOSS version uses a mix of v1.1 and Disney / Pixar training data, tested on MOSS Base. The author states the color tone should be similar to the Krea 2 version. On the same day, the author also released an MOSS-only Euro mixed version.
  • Serves character illustrations and scene images in American comic and Western cartoon styles; the author claims headroom was left for stacking other style or character LoRAs, while trying to avoid identical faces and body types.
  • Full MOSS-specific parameters provided: Strength 0.7-0.85, no trigger word required; 8 step turbo uses 7-9 steps, non-turbo uses 23-27 steps; 10 visible examples, free.

How to Use

  • Load with strength 0.7-0.85, no trigger word needed; run the 8 step accelerated workflow for 7-9 steps, and the standard workflow for 23-27 steps. More steps result in higher saturation.
  • If the style is washed out when stacking character LoRAs: Reduce the character LoRA strength to 0.6-0.85, write fewer photo-related terms, and add 'disney cartoon' or other illustration descriptions.
  • CFG and sampler settings were not provided by the author.

Risks and Limitations

  • The license only allows on-site rental and lending; commercial use of generated images is not permitted.
  • The MOSS version lacks a separate effect comparison description; most style detail descriptions come from the various Krea 2 versions; the page contains mixed content (4 hidden).

Recommended (8)

[A] Aston Martin Dreadnought [Moss] · Moss

LoRA · New Model · Author denrakeiw · Published 11-16 14:50 · Downloads 17

A LoRA for the Aston Martin Dreadnought concept car (a military-grade off-road SUV from a Call of Duty collaboration): Recreates the matte olive green armored body, herringbone carbon fiber, octagonal LED headlight modules, and tactical cockpit screens.

Images

Example 1776x2368

Reasons for Recommendation

  • New model; serves realistic exterior and cockpit images of this concept car: The author claims training used 24 official news images without watermarks (sizes such as 2560x1440), covering underground bunkers, forest wilderness, interior views, and exterior close-ups.
  • Trigger words are registered. The author provided strength settings of 0.85 – 1.0 and prompt writing guidelines: Long natural language paragraphs starting with “Create a photograph of ...”, with labels and screen readings placed in double quotes.
  • The model card includes five complete example prompts (front three-quarters, rear view, driving on mud, cockpit interior) that can be directly adapted; 12 visible example images; free, license allows image generation, sales, and derivatives.

How to Use

  • Write 'Aston Martin Dreadnought Concept' in the prompt, with LoRA strength 0.85 – 1.0; follow the author's examples by writing full descriptions starting with 'Create a photograph of', specifying perspective, scene lighting, and materials.
  • Step counts, CFG, and sampler settings were not provided by the author.

Risks and Limitations

  • The model card header self-identifies as “Flux LoRA” and is organized according to Flux conventions, appearing to reuse the author's Flux template; files are uploaded for MOSS, but recommended sampling parameters for MOSS are not listed.
  • Training annotations contain approximately 165–197 words; fidelity under short prompts is not specified; covers only this specific car, limiting its use case; the page contains mixed content (8 hidden).

Trigger Word: Aston Martin Dreadnought Concept

[A] AG: Bifrons - Fantasy Portraits · MOSS-Image

LoRA · Port · Author Azriaelis · Published 11-16 09:10 · Downloads 41

The MOSS version style LoRA for the Ars Goetia series 'Bifrons': Digital illustration portraits with a fantasy theme, leaning towards a thick paint aesthetic, suitable for DnD-style character avatars.

Images

Example 1024x1536

Reasons for Recommendation

  • Port: The page previously featured Krea 2 and Anima versions; the MOSS version uses the same trigger word @B1FR0NSX. The author provided separate recommended parameters for MOSS (based on Base plus a 8 step acceleration LoRA), though style differences from the other two versions are not documented.
  • Serves fantasy character portraits, tabletop RPG character sheets, and avatar illustrations; the author claims adding the trigger word significantly strengthens the LoRA effect.
  • Full MOSS parameters provided: euler / simple, strength 0.8-1, 8-16 steps, CFG 1-3; 9 visible examples; free.

How to Use

  • Use MOSS Base with a 8 step turbo acceleration LoRA, then apply this LoRA at strength 0.8-1; write @B1FR0NSX in the prompt.
  • Sampler euler, scheduler simple, steps 8-16, CFG 1-3; the specific acceleration LoRA to use was not provided by the author.

Risks and Limitations

  • The author states training images come from mixed sources including Midjourney, NijiJourney, and Tumblr, describing it as a “small experiment with quality varying by LoRA”; material licensing is not disclosed.
  • License only allows image generation and on-site rental; derivatives are not permitted; the page contains mixed content (3 hidden).

Trigger Word: @B1FR0NSX

[A] waxa Medium Format Film Style · v2 Moss Base

LoRA · Version Update · Author superkaero · Published 11-16 10:10 · Downloads 94

Version v2 of the waxa medium format film style photography LoRA: Recreates the wide dynamic range, soft tonal transitions, and shallow depth of field blur of medium format film. This version focuses on compatibility when stacked with character LoRAs.

Images

Example 1376x1824

Reasons for Recommendation

  • Version Update: The author states v2.0 improved compatibility with character LoRAs, at the cost of slightly less soft medium format transitions compared to v1.0—choose v2 if stacking character LoRAs, or stick with v1 for standalone landscape portraits.
  • Serves film-textured portraits, street photography, and realistic landscapes; the author claims training used 98 curated medium format works on Moss Base.
  • Sample image parameters are uniformly public: euler_ancestral / simple / CFG 1 / Steps 8; 12 visible examples; free, license allows image generation, sales, and derivatives.

How to Use

  • Load this LoRA on Moss Base and generate images using the author's sample parameters: euler_ancestral, simple, CFG 1, Steps 8; strength was not provided by the author.
  • v2 has no registered trigger word (v1 registered 'medium format film'); you can write medium format film-related descriptions in the prompt.

Risks and Limitations

  • Sample images were generated on Base with 8 steps and CFG 1; it is not stated whether an acceleration LoRA was also used, so confirm your own workflow before copying parameters.
  • The author admits v2's soft transitions are inferior to v1; the page contains mixed content (8 hidden).

[A] The Layout Studio for MOSS-Image - "Draw" your bbox prompt, then paste the JSON in any generator · v1.1.0

Other · Version Update · Author Rexoplanets · Published 11-16 19:54 · Downloads 6

Layout Studio v1.1.0: An offline tool that allows you to draw MOSS bbox layouts by dragging boxes in a browser and export JSON prompts. This version adds the ability to import existing bbox prompts from PNG metadata.

Images

Example 1728x1152

Reasons for Recommendation

  • Version Update: Added PNG metadata import, allowing you to drag images generated by ComfyUI, Civitai online generators, or Tensor Art into the tool to directly restore the MOSS bbox JSON and modify the layout; the author states PNGs from these three sources have been verified.
  • Serves scenarios requiring "who goes where" such as preventing character crossover (mixing clothing/features), poster text positioning, product image composition, and storyboarding: Drag boxes on the canvas, and the tool automatically converts them to MOSS yxyx coordinates (-500 to 500, with the origin at the center).
  • Single HTML file; the author claims it is completely local and offline, independent of ComfyUI or Python; exported JSON can be pasted into any generator running MOSS; free and open source.

How to Use

  • Unzip the attached zip and open the HTML in a browser; after dragging boxes to layout, copy the JSON and paste it into the prompt box of the MOSS generator; you can also export plain text prompts for models that do not recognize JSON.
  • The author advises keeping CFG ≥ 2.5, otherwise elements will drift; official recommendations are ≥ 50 steps, 1.5 MP or 2 MP, and 3-6 CFG.
  • Do not write absolute positions like "left / right" in the description of each element, as text descriptions will override the box positions.

Risks and Limitations

  • Not model weights; does not change MOSS capabilities; the author admits MOSS does not strictly adhere to boxes, and text element positioning is more accurate than objects and characters.
  • Canvas and box sizes are aligned to multiples of 16 pixels, which is a MOSS requirement; custom sizes will be adjusted; the tool has not been runtime-verified.

[A] Mobile Cell Phone Photography · MOSS 2.0

LoRA · Port · Author FuenfFingerFaultier · Published 11-16 16:00 · Downloads 73

MOSS version of the casual smartphone snapshot style LoRA: Mimics uncomposed phone photos, featuring selfie and close-up wide-angle distortion with sharp backgrounds (no bokeh).

Image

Example 768x1360

Reasons to Recommend

  • Port: The same page previously featured versions for SDXL, Flux.1, ZImageTurbo, Klein 4B, Krea 2, etc.; the author states that MOSS 2.0 was retrained using the same data as before, with differences not specified
  • Suitable for realistic images needing a "lived-in feel" rather than studio polish: selfies, social media-style photos, and wide-angle close-ups can help suppress the refined shallow depth of field common in AI-generated images
  • Trigger words photo and selfie are registered; 19 visible examples make this one of the most example-rich entries this period; free

How to Use

  • Add photo to the prompt, and add selfie for selfie scenarios
  • The author did not provide strength, steps, CFG, or sampler settings

Risks and Limitations

  • The description is only two sentences long, with no MOSS parameters provided; in the Krea 2 version, the author noted a tendency toward a painterly style requiring prompt correction, but it is unclear if the MOSS version shares this issue
  • The license only allows on-site rental and lending, not commercial use of generated images; high download counts for the entire model page only indicate popularity

Trigger Words: photo, selfie

【A】MOSS-Image SFT Mixed INT8/W4 Quants · 3.7 GB

Base Model / Checkpoint · New Model · Author BakaPotatoLord · Published 11-16 18:06 · Downloads 21

MOSS-Image SFT INT8 / INT4 mixed quantized base model: Approximately 3.72 GiB, designed for machines with limited VRAM and RAM; the author successfully ran it on a GTX 1660 Super with 6 GB.

Image

Example 688x1040

Reasons to Recommend

  • New model; the author tested precision layer by layer against the official INT8 version: 107 layers use INT4 (W4A4 ConvRot MSE), while 61 layers retain the official INT8 ConvRot weights
  • Enables MOSS text-to-image generation on low-spec GPUs: The author's test environment was a GTX 1660 Super with 32 GB RAM, achieving approximately 5.7 s/it at 696 × 1048 resolution
  • Test parameters are public: 8/12 steps, CFG 1, Euler / Simple; the example workflow includes Lodestone's 8 step acceleration LoRA; free, with a license allowing image generation, sales, and derivatives

How to Use

  • Replace the MOSS SFT base model load; follow the author's settings of 8/12 steps, CFG 1, Euler / Simple, and pair with Lodestone's 8 step LoRA
  • The specific nodes or ComfyUI versions required to load INT8 / W4A4 ConvRot weights are not provided by the author; refer to the workflow in the example image

Risks and Limitations

  • Only 1 visible examples (2 hidden), with no quality comparison against the official INT8 or bf16 versions
  • This is a quantization of SFT; it is unclear if LoRAs trained for Base will yield consistent results when applied here; speed remains slow on low-spec machines

【A】2003 China Summer · MOSS

LoRA · Port · Author 858654128784 · Published 11-16 10:47 · Downloads 91

MOSS version of the 2003 China Summer concept LoRA: Renders images as casual snapshots from Chinese summers around 2008 years ago—dappled tree shade light, faded warm tones, slight highlight bloom, accompanied by bamboo chairs, palm-leaf fans, popsicles, convenience store freezers, iron bunk beds, old CRT monitors, and bus interiors, resembling photos developed from old compact cameras.

Generating one image with and one without the LoRA using the same prompt makes it easiest to see which era-specific props and color tones were added.

Image

Example 896x1152

Reasons to Recommend

  • Port: The same page previously featured a Krea 2 version; the MOSS version uses the same description, with differences between versions not specified by the author; the filename is chinasummer2008, but the model name says 2003, so rely on the "around 2008 years ago" mentioned in the description
  • Suitable for nostalgic text-to-image themes: The author states it works for portraits, street scenes, interiors, and still life, with characters typically wearing summer clothes like short sleeves, shorts, and sandals
  • The author provided a prompt structure that is easy to follow; 3 visible examples (5 hidden), free, with a license allowing image generation, sales, and derivatives without requiring attribution

How to Use

  • Start the prompt with hazy summer snapshot, then describe the character's appearance and clothing, actions and position, background, and end with a sentence about lighting; the author's example is "Soft dappled daylight and faded warm colors give a nostalgic compact-camera appearance."
  • No trigger words are registered; LoRA strength, steps, CFG, and whether to use Base or SFT were not provided by the author

Risks and Limitations

  • No strength recommendations are given, requiring users to experiment starting from medium strength; the prompt writing style in the description applies to the entire model page, and it is unclear if adjustments are needed for the MOSS version
  • The page contains mixed content, with 5 examples hidden; there is an inconsistency between the year in the name and the year in the description, which the author did not explain

【A】Nix Comic [Multiple Styles] · Euro [MOSS]

LoRA · Version Update · Author xFennec777 · Published 11-16 20:48 · Downloads 4

Euro [MOSS] variant of the Nix Comic style LoRA: Mixes European comic art styles into Nix Comic's flat-color base, featuring flat color blocks with rough, dynamic lines and hatching; currently only available for MOSS.

Image

Example 1248x1680

Reasons to Recommend

  • Version update but not an upgrade; it is a new art style added to the same page: The author states Euro is created by mixing NixComic MOSS with three style LoRAs—an unreleased Massimo Fecchi style (defining core style), Tintin style (flattening, coloring), and Yevgeniy Migunov style (lines and hatching); a Disney / Comic [MOSS] version was also released on the same day
  • Suitable for European comic illustrations from Belgium, Italy, to Russia: The author claims it helps MOSS approach the styles of artists like Watterson, Franquin, Migunov, and Hergé, and serves well as a base for LoRA mixing
  • Parameters are fully provided: The model description specifies 0.7-0.85 strength for the Nix Comic series with no keywords required; for the MOSS version, use 7-9 steps for 8 step turbo and 23-27 steps for non-turbo 25 steps; 9 visible examples, free

How to Use

  • Load and use immediately with no trigger words; start strength at 0.7-0.85, use 7-9 steps for turbo sampling and 23-27 steps for non-turbo; excessive steps will cause oversaturation
  • When stacking with strong character LoRAs, the art style may be suppressed: Reduce the character LoRA to 0.6-0.85, minimize photo-type words, and add disney cartoon or desired illustration descriptions

Risks and Limitations

  • The author states this is not their usual style; v1 has style drift, the file size of approximately 315mb is large, with plans to reduce it in v2; the 0.7-0.85 strength recommendation applies to the entire series and was not specified separately for the Euro version
  • It mixes LoRAs imitating named artists and works, one of which is unpublished; the license only allows on-site rental and lending, excluding commercial use of generated images; the page contains mixed content with 11 examples hidden, and it was published less than 1 hours ago

Worth Watching (7)

【B】Polestar Formula 2030 [Moss] · v1

LoRA · New Model · Author denrakeiw · Published 11-16 20:41 · Downloads 3

Polestar Formula 2030 concept car model LoRA: An ice-blue four-door shooting brake sports car, recreating the B-pillar-less suicide doors, dual-blade LED light clusters, and 2+2 minimalist cockpit.

Image

Example 1776x2368

【B】Style: Terada Katsuya / Rakugakingu · MOSS-base v1.0

LoRA · New Model · Author NerissG · Published 11-16 20:00 · Downloads 9

Terada Katsuya (Rakugakingu) style LoRA: A sketch illustration style trained exclusively on works he created for "The Legend of Zelda: A Link to the Past."

Image

Example 832x1248

【B】Fusheng Yunhui style | MOSS Image LoRA · v1.0

LoRA · New Model · Author NO83 · Published 11-16 19:19 · Downloads 26

Fusheng Yunhui style LoRA: Primarily learns composition and lighting—subject placement, foreground/background layering, and using light and color to guide the eye toward the subject, leaning towards Eastern decorative illustration.

Image

Example 1360x2048

【B】Moss Japanese Anime Thin Paint · v1.0

LoCon · New Model · Author XiaoXiaoQAQ · Published 11-16 14:17 · Downloads 44

Moss Japanese Anime Thin Paint: A LoCon for the Japanese anime "thin paint" style, which is a lightweight cel-shaded illustration style with light coloring and few layers.

Image

Example 832x1200

【B】MOSS-Image Pose Control ComfyUI | DWPose + Edit LoRA · v1.0

Workflow · New Model · Author BearGhasemian · Published 11-16 11:12 · Downloads 26

MOSS-Image Pose Control ComfyUI workflow: Uses DWPose to extract skeletons from reference images, then combines pose control LoRAs and MOSS-Image Edit text encoding to allow new images to retain the reference pose while changing characters, clothing, and scenes.

Image

Diagram 1086x1448

【B】MOSS-Image Canny Control ComfyUI | Edge-Guided Restyling · v1.0

Workflow · New Model · Author BearGhasemian · Published 11-16 11:13 · Downloads 20

MOSS-Image Canny Edge Control ComfyUI workflow: Extracts Canny edges from photos as structural constraints, preserving main outlines while using prompts to change materials, colors, and lighting for restyling products, vehicles, and buildings.

Image

Video 1024x1024

【B】Russian Fashion from 90s by Dizzy · v1

LoRA · New Model · Author DizzyHarvester511 · Published 11-16 16:51 · Downloads 0

Russian Fashion from the 90s: A LoRA for Russian street fashion style from the 90s / Perestroika era, dressing characters in outfits from that period.

Image

Example 1024x1024

Included Only (3)

  • Bemi Style — NAI5 · v1.0 — LoRA: Bemi Style: A MOSS style LoRA that replicates the art style of artist Bemi on NAI5. The author states they chose it because the overall image is clean and refreshing. Illustration 1672x941
  • hisako_kotobuki (seitokai nimo ana ha aru) · v1.0 (moss) — LoRA: A MOSS version character LoRA for hisako kotobuki (brown hair, yellow eyes) from the manga Seitokai ni mo Ana wa Aru. The author describes it as a casual "zatsu lora" (practice LoRA). Illustration 1448x1086
  • KOREAN COSPLAY BY STX (moss) · X1 — LoRA: KOREAN COSPLAY BY STX: A paid Korean female cosplay concept LoRA, used for generating cosplay-style portrait photos. Example 768x1344

Data: Civitai public API (base model MOSS Image), inclusion window 11-16 08:31 → 11-16 21:01 (Beijing Time). Recommendations have been reviewed by editors; download counts only reflect popularity.

Civitai

Explore the latest AI Radar updates