MOSS Image Evening Report 2026-11-12 | Top Pick: [Style] dAIver's Inkveil for MOSS Image

This issue reviews 32 new MOSS versions, with 3 S-tier, 5 A-tier, 12 B-tier, and 12 C-tier entries. Top pick: Inkveil, a dark fantasy night-scene LoRA featuring liquid ink brushstrokes and glowing filigree patterns, with trigger strings and strength settings provided. ElectroPopPoster (neon high-saturation poster re-rendering, with complete strength tiers and sampling recipes) and HaDeS 8-step Turbo v1.0 (reducing facial bias) are both S-tier; Cursed VHS liminal horror style, realistic night streetlights, Chinese rural wall art, Fluorescentizer fluorescent style, and others are A-tier. Tile Control beta3 has detailed documentation but no visible examples, earning a B-tier rating, while multiple standard character LoRAs and workflows with only a single sentence of description are rated C-tier.

Zhenzhen's AI Radar · MOSS Image · 2026-11-12 Evening Report

This issue features 30 new versions: Strongly Recommended 3 · Recommended 5 · Worth Watching 11 · Listed Only 11

Editor's Note This issue reviews 32 new MOSS versions, with 3 S-tier, 5 A-tier, 12 B-tier, and 12 C-tier entries. Top pick: Inkveil, a dark fantasy night-scene LoRA featuring liquid ink brushstrokes and glowing filigree patterns, with trigger strings and strength settings provided. ElectroPopPoster (neon high-saturation poster re-rendering, with complete strength tiers and sampling recipes) and HaDeS 8-step Turbo v1.0 (reducing facial bias) are both S-tier; Cursed VHS liminal horror style, realistic night streetlights, Chinese rural wall art, Fluorescentizer fluorescent style, and others are A-tier. Tile Control beta3 has detailed documentation but no visible examples, earning a B-tier rating, while multiple standard character LoRAs and workflows with only a single sentence of description are rated C-tier.

Strongly Recommended (3)

[S] [Style] dAIver's Inkveil for MOSS Image · 1.0 (MOSS)

LoRA · New Model · Author Experimental_dAIver · Released 11-12 15:49 · Downloads 32

Inkveil style LoRA: liquid ink brushstrokes + glowing filigree patterns + soft painterly edges, creating an overall quiet, dark fantasy night scene.

Images

Example 1992x3154

Reasons for Recommendation

  • New model; training materials include night scenes, elves, animals, and quiet figures, suitable for dark fairy-tale illustrations, nocturnal animals, and spirit/mage character images; the author states that faces remain recognizable at normal strength, large areas of negative space remain soft, and strong rim lighting and translucent surfaces best highlight the filigree patterns
  • Usage is fully documented: trigger string inkveil, liquid ink strokes, glowing filigree, soft painterly edges, ethereal dark atmosphere, strength 0.55–0.75; the model card includes full example prompts for scenes like a black cat in a flower field, cat silhouette landscapes, spirits in ruins, and a girl surrounded by elves
  • 15 visible example images, free to download

How to Use

  • Start the prompt with the full trigger string, then describe the subject, glow color (e.g., cyan filigree), and night environment; strength 0.55–0.75
  • Steps, CFG, and sampler are not specified by the author

Risks and Limitations

  • The author warns that when strength exceeds 0.8, close-ups may drift toward the training set subjects; strong daylight scenes and pure comic line art are weak points
  • 4 examples are hidden; the license does not allow derivatives

Trigger Words: inkveil, liquid ink strokes, glowing filigree, soft painterly edges, ethereal dark atmosphere

[S] ElectroPopPoster · ElectroPopPoster_v1_e5

LoRA · New Model · Author Cosmic_Crafter · Released 11-12 15:49 · Downloads 13

ElectroPop Poster style LoRA: re-renders any subject into a neon high-saturation poster illustration with glowing outlines and dense color blocks; intensity is adjustable from subtle color grading to psychedelic.

A lighter version from epoch 3 is available on the same page; choose this if you prefer a more blended, less intense effect.

Images

Example 1024x1024

Reasons for Recommendation

  • New model; suitable for high-saturation illustration of posters, covers, portraits/vehicles/food/mecha. The author states that in 39 extreme tests: when prompts specify other media (charcoal, watercolor, ukiyo-e, pixel art), the LoRA yields and amplifies that medium, only taking over prompts without specified media
  • Strength tiers are clearly documented: 0.75 as a color grading layer, 1.0 for full style, 1.3 for maximum saturation, and maximum 2.5 for full psychedelic effect; the author's test recipe uses euler / sgm_uniform, approximately 50 steps, CFG 3.0
  • The author claims that required titles and sign text render clearly at 1.0; 8 visible example images, free with a license allowing commercial use

How to Use

  • Place pntrcr at the beginning of the prompt (it works even if omitted, as the style is bound to the weights); choose strength 0.75 / 1.0 / 1.3 as needed, or lower to 0.5–0.75 for a restrained color grading effect
  • Use euler / sgm_uniform, approximately 50 steps, CFG 3.0

Risks and Limitations

  • The style is very heavy at 1.0; photorealistic requests may still yield a photographic feel; complex scenes occasionally produce meaningless graffiti text
  • The training set contains only 75 images over 5 epochs, trained using the site's trainer; approximately 50 steps is slow, and the author has not commented on performance with acceleration LoRAs

Trigger Words: pntrcr

[S] HaDeS MOSS Turbo · HaDeS v1.0 Turbo 8Steps

LoRA · Version Update · Author _HaDeS_ · Released 11-12 16:07 · Downloads 74

HaDeS MOSS Turbo 8-step v1.0: a 8-step acceleration LoRA for MOSS-Image Base, with a file size of only 3.56 MB; this version focuses on reducing facial bias.

Suitable for comparing the same prompt with CFG 1 / CFG 1.5 to see two different flavors.

Images

Example 1960x1394

Reasons for Recommendation

  • Version update: the author documented changes—the old turbo version tended to generate East Asian faces when subjects were undescribed, while v1.0 aims to preserve Base diversity; according to the author's statistics (same prompt, 16 seeds), Base 25 steps yielded approximately 2-4 images, the previous 8-step version yielded approximately 11-12 images, and v1.0 yielded approximately 8 images, showing reduced but not eliminated bias
  • This version specifies the base model as MOSS-Image Base; rank 2, acting only on 24-41 layers, trained using the HaDeS MOSS Suite's built-in script without an image dataset
  • Two output flavors: CFG 1 is lighter and softer; CFG 1.5 has higher contrast, color, and sharpness but runs twice per step; the author claims that signs, three-line posters, and other text in the test set render correctly at both CFG levels; 4 visible examples, none hidden

How to Use

  • Base model moss_image_base_bf16 (Comfy-Org/MOSS-Image), paired with Qwen3.5-9B text encoder and FLUX.2 VAE
  • Native nodes: LoraLoaderModelOnly → KSampler, 8 steps, CFG 1 or 1.5, euler / simple, strength 1.0; or use the MOSS Turbo LoRA node from HaDeS MOSS Suite connected to the turbo input of the MOSS KSampler
  • Specify character features in the prompt if a specific appearance is desired

Risks and Limitations

  • Other strengths/steps, sizes above 1536 px, img2img, and editing have not been tested by the author; the author claims good results on SFT but has not quantified them
  • Facial bias is only reduced; the Civitai page license does not allow derivatives, while the model card states Apache 2.0, so refer to the author's Hugging Face page for the authoritative license

Recommended (5)

[A] DC Cursed VHS LiminalCore · v1.0 [MossBase]

LoRA · New Model · Author dkpc69 · Released 11-12 20:58 · Downloads 5

Cursed VHS LiminalCore style LoRA: mimics eerie VHS recording frames—faded colors, soft focus, deep shadows—creating a nightmarish horror/fantasy feel for familiar places and strange figures.

Downgraded from automatic S to A: prompt writing is very complete, but strength and sampling parameters are not listed on the page.

Images

Example 1088x1920

Reasons for Recommendation

  • New model; suitable for liminal spaces, horror atmospheres, and surreal dreamscapes; the author states it mimics the old recording style of TikTok account @deadtempovisions
  • Training information is public: trained on MOSS Base for 7000 steps, Rank 32, GPT4o tagging, 726 image dataset; the author states it also works normally on MOSS SFT
  • Prompt structure provided: trigger word → subject appearance/action → main light source → environment/composition → tone/atmosphere → overall lighting ratio → desired viewer emotion; auxiliary words like eerie, liminal are also listed; 13 visible example images, the author claims all examples include prompts and ComfyUI workflows

How to Use

  • Start the prompt with DCVHSLMNLCR, , write complete sentences following the author's structure, and add words like eerie, surreal, dreamlike, unsettling, muted
  • Strength, steps, and CFG are not provided by the author; check the workflows attached to the example images

Risks and Limitations

  • Mimics the visual style of a specific creator; 7 examples are hidden, and the page contains mixed content
  • Just released, with almost no download feedback

Trigger Words: DCVHSLMNLCR,

[A] ElectroPopPoster · ElectroPopPoster__lighter

LoRA · Version Update · Author Cosmic_Crafter · Released 11-12 15:52 · Downloads 27

Lighter version of ElectroPop Poster: an epoch 3 checkpoint from the same training run, which the author states is more blended and less intense than the main version (epoch 5).

Images

Example 1024x1024

Reasons for Recommendation

  • Version update: the update note is only one sentence—epoch 3 of the same training, yielding more blended images; all other descriptions and parameters are shared with the main version
  • Suitable for scenarios where you only want neon poster color grading without letting the style completely take over the image; 5 visible example images

How to Use

  • Follow the main version's parameters: place pntrcr at the beginning of the prompt (this version does not register a trigger word, but the author states it works even if omitted); strength 0.75 / 1.0 / 1.3; euler / sgm_uniform, approximately 50 steps, CFG 3.0

Risks and Limitations

  • The difference from the main version is described in only one sentence, with no comparison images; page parameters are written based on the main version

[A] Realistic Night Outdoor Photo — Real Streetlit Night Scenes with Clean Shadows | MOSS · MOSS

LoRA · New Model · Author 858654128784 · Released 11-12 16:17 · Downloads 74

Realistic night outdoor photo LoRA: pulls the base model's gray or overexposed night scenes back to realistic night colors—deep blue skies, dominated by streetlights and shop lights, with faces and clothes still clearly visible, and clean, non-muddy shadows.

Images

Example 896x1152

Reasons for Recommendation

  • New model; suitable for night street portraits/scenes like bus stops, street stalls, canal railings, parking lots, rooftops, and alleys
  • Prompt writing is clearly documented: start with nightoutdoorphoto, , describe the person, clothing, location, and action in the middle, and end with outdoors at night; the author states that training annotations were very short, so long prompts are fine, as the night atmosphere is handled by the LoRA
  • Free, license allows commercial use; the same author released a "Chinese rural wall art" concept LoRA on the same day

How to Use

  • nightoutdoorphoto, + person/clothing/location/action + outdoors at night
  • Strength, steps, CFG not provided by author

Risks and Limitations

  • Only 3 visible examples, 5 hidden, page contains mixed content
  • No parameter suggestions, compatibility with acceleration LoRAs not specified

[A] Rural Wall Print · MOSS

LoRA · New Model · Author 858654128784 · Published 11-12 09:54 · Downloads 73

A concept LoRA for "Chinese rural wall prints": generates an image of a printed poster pinned to a peeling gray wall, with creases and curled edges, surrounded by typical household items like windows, cabinets, thermos flasks, and fruit platters.

Images

Example 896x1152

Reasons for Recommendation

  • New model; suitable for scene photos with a sense of era or rural themes: the content of the print can be freely replaced, with the author citing examples like the Matsuyama Waterfall with a red sun, peonies and birds, cranes, koi, galloping horses, and busy farming scenes
  • Three-part prompt structure: starts with "rural wall print, a photograph of a paper illustration fastened to an aged plaster wall.", then "The printed artwork depicts ..." to describe the content of the print, and finally "In the actual room, ..." to describe the room environment
  • Free, license allows commercial use; 73 downloads since release, high attention in this issue

How to Use

  • Write prompts following the author's three-part structure: prefix sentence → The printed artwork depicts ... → In the actual room, ...
  • Strength, steps, CFG not provided by author

Risks and Limitations

  • Only 3 visible examples, 5 hidden, page contains mixed content
  • No parameter suggestions

[A] Fluorescentizer · ⚪ Moss

LoRA · Ported · Author Lara_De_Martin · Published 11-12 21:00 · Downloads 2

The MOSS version of Fluorescentizer: generates images with fluorescent colors and strong contrast, with a slight anime feel, focusing on glowing effects in dark areas.

Upgraded from automatic B to A: clear purpose and trigger words, many visible examples, only missing MOSS-specific parameters.

Images

Example 1536x1408

Reasons for Recommendation

  • Ported (same page already has Krea 2 A/B, Flux.1 A/B versions); the MOSS version has only one file, with no separate description
  • The author states it works best in dark scenes; prompts for daytime/bright scenes still produce vibrant fluorescent colors, but contrast is reduced
  • Registered trigger words: fluorescentizer, fluorescent colors, neon colors, glowing colors, anime; the author recommends stacking tags like DIGITAL ILLUSTRATION; 12 visible example images

How to Use

  • Add fluorescentizer to the prompt, then add digital illustration / digital art; to enhance, add fluo color, fluorescent, glow, by night, dark background, psychedelic background, or add anime or manga
  • Strength, steps, CFG not provided by author

Risks and Limitations

  • The model card lists two optional A/B versions, but the MOSS version has only one file, and it is not specified which version it is; 8 examples are hidden, page contains mixed content
  • License does not allow derivatives; just released, the MOSS version has almost no download feedback yet

Trigger Words: fluorescentizer, fluorescent colors, neon colors, glowing colors, anime

Worth Watching (11)

[B] xUnderdark · MOSS Image

LoRA · New Model · Author randomimagepotato · Published 11-12 19:20 · Downloads 20

xUnderdark style LoRA: described by the author in one sentence as "Drow-style Underdark," i.e., a dark fantasy underground world theme.

Images

Example 1040x1520

[B] Moss-Image Tile Control · beta3

ControlNet · Version Update · Author Xabsurd · Published 11-12 13:22 · Downloads 27

MOSS-Image Tile Control beta3: Tile control for MOSS, preserving the original image's composition, subject outlines, and local structures, allowing the model to redraw texture details; this version allows fine-tuning the degree of change with a single strength parameter.

Images

Diagram 1254x1254

[B] Psychedelic Poster · Moss

LoRA · Ported · Author CN_Signal · Published 11-12 13:52 · Downloads 33

The MOSS version of Psychedelic Poster: a psychedelic poster style LoRA.

Images

Example 1280x1920

[B] Vetehine ┃ Hellgasm · v.0.1

Base Model / Checkpoint · New Model · Author Vetehine · Published 11-12 20:19 · Downloads 0 · Paid

Hellgasm v0.1: a paid MOSS full base model, focusing on gothic dark fantasy, horror, and surrealism, while also accommodating cinematic realism and anime illustration.

Images

Example 896x1152

[B] Manga style (Berserk) · Moss v1

LoRA · Ported · Author Yofaraway · Published 11-12 16:25 · Downloads 1 · Paid

The MOSS version of the "Berserk" manga style LoRA: a black-and-white manga style in the manner of Kentaro Miura.

Images

Example 1024x1536

[B] VÆLORIA_moss-image · v1.0

LoRA · New Model · Author Vince_AI · Published 11-12 14:31 · Downloads 2 · Paid

VÆLORIA style LoRA: surreal + high-gloss fashion editorial style, featuring exaggerated makeup, exotic materials, floating objects, and sculptural clothing.

Images

Example 832x1216

[B] Kigurumi Onesie: Sheep [Krea2/MOSS] · MOSS_v1.0

LoRA · Ported · Author AuthentiQ · Published 11-12 09:35 · Downloads 21

The MOSS version of the Sheep Kigurumi Onesie clothing LoRA: puts a sheep-shaped onesie on the character.

Images

Example 815x1448

[B] Moss-Image Tile Control · beta2

ControlNet · New Model · Author Xabsurd · Published 11-12 13:19 · Downloads 7

MOSS-Image Tile Control beta2: Uses the built-in MOSS-Image Color/Depth Control nodes in ComfyUI plus a Tile LoRA for Tile control, without needing to install custom nodes.

Images

Diagram 1254x1254

[B] Mochizuki Kei Style 【krea2&MOSS&anima】 · MOSS-heal v1.0

LoRA · Ported · Author NXMZ · Published 11-12 20:04 · Downloads 17

A LoRA imitating the illustration style of Mochizuki Kei, trained on KBlueLeaf's MOSS-Image-heal.

Images

Example 1280x1856

[B] Comfyui simple MOSS workflow with guide · v1.0

Workflow · New Model · Author PotatCat · Published 11-12 15:18 · Downloads 31

PotatCat's simple text-to-image ComfyUI workflow for MOSS, accompanied by a usage guide article.

Images

Example 1344x1728

[B] Body weight slider. (Skinny ===> Fat slider) · Moss-Image

LoRA · Ported · Author Maruhashi · Published 11-12 20:19 · Downloads 1 · Paid

The MOSS version of the body weight slider LoRA: negative values make the subject thinner, positive values make them fatter, no trigger word required.

Images

Diagram 1672x941

Included Only (11)


Data: Civitai Public API (base model MOSS Image), inclusion window 11-12 09:03 → 11-12 21:01 (Beijing Time). Recommendations are editorially reviewed; download counts only reflect attention.

Civitai

Explore the latest AI Radar updates