MOSS Image Evening Report 2026-11-02 | Top Pick: Motion_rush

Zhenzhen's AI Radar · MOSS Image · 2026-11-02 Evening Report

This issue features 31 new versions: Strongly Recommended 0 · Recommended 6 · Worth Watching 10 · Included Only 15

Editor's Note This issue reviews 35 new MOSS versions. Top pick: Motion_rush, which turns the panning technique from photography into a LoRA, keeping the subject sharp while streaking the background along the direction of motion; its prompt structure is copy-paste ready and the license is fully open. Four other layout/lens LoRAs by the same author—Ceiling Fisheye Surveillance, DVD Back Cover, 0.5x Ultra-Wide Selfie, and Full-Cover VHS Rental Store—along with Fwaun Style v5.1, are all rated A. Love and Pop cinematic photography style is downgraded from S to B; Azriaelis’s three style transfers are downgraded from S to B/C; Mr Piglets released four character versions of Lux43–46 within the same hour, so only Lux46 is retained while the others are excluded; numerous standard character and person LoRAs have insufficient documentation and are rated C.

Recommended (6)

[A] Motion_rush · v1.0

LoRA · New Model · Author Reversoma · Released 11-02 17:50 · Downloads 29

A LoRA for the "panning" effect: keeps the subject sharp while applying strong directional motion blur to the background, used to generate cinematic action shots of people, animals, or vehicles moving at high speed.

Maintained as A: The effect is clearly described in one sentence, the prompt structure is copy-paste ready, and the license is maximally open. However, it lacks sampling parameters and has only been tested on third-party Turbo models, so it does not reach S.

Images

Reasons for Recommendation

• New model: Turns the panning technique from photography into a LoRA; the author states it was trained on realistic photos of moving people, animals, and vehicles.
• Serves realistic scenarios requiring a sense of speed where "the subject is frozen while the background flies by," such as sports, racing, running animals, or street action shots; the author provides prompt structures, not just a trigger word.
• 7 visible examples; free with a fully open license (including merging and selling merged models); approximately 3 hours since release with 29 downloads (indicating attention only).

How to Use

• Load motion_rush_moss_v1.safetensors (69 MB) and start the prompt with "motion_rush".
• Describe the subject as "in sharp focus" and the environment as "strong directional motion blur in the background".
• The author tested this on Kromoss Turbo with strength 0.8 and 1.0; steps, CFG, and sampler were not provided by the author.

Risks and Limitations

• The author states that complex angles, multiple subjects, and anime styles are still experimental; composition, subject clarity, and orientation may be unstable.
• The base model used for testing was Kromoss Turbo (not official MOSS); performance on the official Base / SFT models is not specified by the author.
• 3 sensitive-level examples are hidden; this entry has not been run.

Trigger Words: motion_rush, in sharp focus

[A] Ceiling Fisheye Cam — Straight-Down 360° Security View, the Room as a Floor Plan | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-02 19:12 · Downloads 35

A LoRA for a ceiling-mounted fisheye surveillance perspective: circular frame with black borders, a line of white text at the bottom showing camera name and timestamp; the room appears as a "floor plan," furniture shows only top surfaces, walls and doors/windows fall outward toward the rim, people appear only as heads and backs, and night mode can switch to infrared black and white.

Maintained as A: The effect is unique, and the lens geometry and prompt structure are clearly written for replication. However, it lacks strength and sampling parameters, and visible examples are fewer than hidden ones, so it does not reach S.

Images

Reasons for Recommendation

• New model: Turns the specific visual of "ceiling-mounted fisheye surveillance facing straight down" into a LoRA; the author clearly explains the lens geometry: furniture viewed from above, walls falling outward at the rim, and people with strong perspective foreshortening.
• Serves narrative images in surveillance screenshot style, suspense scenes, or overhead layout diagrams of dorms or rooms; the author provides complete prompt structures and tag examples (e.g., CAM 04 DORM 312 plus timestamp).
• 4 visible examples; free with a fully open license; approximately 2 hours since release with 35 downloads (indicating attention only).

How to Use

• Load ceilingfisheyecam.safetensors (142 MB); the author did not register a trigger word, but the description states to start with "ceilingfisheyecam, overhead circular fisheye view of".
• Next, describe the room, then explicitly state that the camera is in the center of the ceiling facing straight down: furniture seen from the top, walls falling outward at the rim, people seen as heads and backs; then describe who is doing what where, followed by the camera name and timestamp, and finally specify color or infrared black and white.
• The author notes that omitting the "facing straight down" part often results in a corner angle shot with a circular frame added.
• Strength, steps, CFG, and sampler were not provided by the author.

Risks and Limitations

• No strength or sampling parameters are provided; the trigger word is not registered in the version info but only appears in the description text.
• 14 sensitive-level examples are hidden, far more than the 4 visible examples.
• This entry has not been run.

[A] DVD Back Cover — Hero Still, Sample Stills, Synopsis, Barcode | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-02 19:09 · Downloads 28

A LoRA for DVD back cover layout: generates the entire back of a DVD case in one image, featuring a large hero still at the top, a row of small sample stills below, a synopsis, cast and crew list, runtime/rating/audio track spec bar, and a barcode at the bottom.

Maintained as A: The purpose is clear, the prompt order is copy-paste ready, and it highlights the use of MOSS layout mode; however, it lacks strength and sampling parameters, and visible examples are few.

Images

Reasons for Recommendation

• New model: Turns the multi-element layout of a "DVD back cover" into a LoRA; the author states it can be applied to genres like sports films, late-night convenience store thrillers, road movies, and documentaries.
• Serves film poster design, fictional movie materials, or fan-made "fake movie" covers; the author notes that using MOSS layout mode (JSON with a box for each element) makes the layout more stable.
• 4 visible examples; free with a fully open license; approximately 2 hours since release with 28 downloads (indicating attention only).

How to Use

• Load dvdbacklayout.safetensors (142 MB); the author did not register a trigger word, but the description states to start with "dvdbacklayout, DVD back cover".
• Sequentially write the genre, hero still (who, where, doing what), sample stills (one thing per frame), synopsis, cast and crew list, spec bar, and barcode, ending with the color scheme.
• Text in the image should be only in English or Chinese; the author notes that Japanese will result in garbled characters; for more stable layouts, switch to MOSS layout mode.
• Strength, steps, CFG, and sampler were not provided by the author.

Risks and Limitations

• No strength or sampling parameters are provided; the trigger word is not registered in the version info but only appears in the description text.
• 16 sensitive-level examples are hidden, far more than the 4 visible examples; the author notes that placing multiple people in small frames will cause them to blur together.
• This entry has not been run, and the layout mode prompt structure has not been verified.

[A] 0.5x Ultra-Wide Selfie — Sole in Your Face, Big Head Small Body, Bent Edges | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Released 11-02 19:04 · Downloads 40

A LoRA for 0.5x ultra-wide phone selfies: hands, faces, or shoe soles close to the lens are magnified several times, the body behind shrinks rapidly, room corners, bed edges, and subway handrails bend outward, and the entire space is squeezed into one frame.

Maintained as A: The perspective effect is described specifically, and both prompt order and pitfalls are provided; however, it lacks strength and sampling parameters, and examples are primarily hidden sensitive-level images.

Images

Reasons for Recommendation

• New model: Turns the perspective distortion of close-range ultra-wide selfies into a LoRA; the author incorporates the rule "whatever is closer to the lens is magnified" into the prompt structure.
• Serves quirky selfies, trendy portraits, or exaggerated perspective portraits from high/low angles; the author provides camera position descriptions such as phone placement (held high, on the ground, or extended over a railing).
• 4 visible examples; free with a fully open license; approximately 2 hours since release with 40 downloads (indicating attention only).

How to Use

• Load ultrawideselfie.safetensors (142 MB); the author did not register a trigger word, but the description states to start with "ultrawideselfie,".
• First describe the phone position, then the body part closest to the lens and magnified (face, hand, shoe sole), followed by how the body shrinks and how the room bends into the frame, ending with the character's appearance, age, and expression.
• Explicitly state the number and position of legs and feet (e.g., exactly two legs), otherwise the wide angle may generate an extra leg; keep the face about one arm's length from the lens, as getting closer will cause blurring.
• Strength, steps, CFG, and sampler were not provided by the author.

Risks and Limitations

• No strength or sampling parameters are provided; the trigger word is not registered in the version info but only appears in the description text.
• 10 sensitive-level examples are hidden, more than the 4 visible examples, and the description includes prompts for adult content.
• Wide angles may generate extra limbs, and close range may cause distortion; the author advises constraining this in the prompt; this entry has not been run.

[A] Fwaun Style · moss_image_v5.1

LoRA · Version Update · Author fwaunstp733 · Released 11-02 10:57 · Downloads 51

The second MOSS version of the Fwaun Style LoRA: a flat-color, line-art-free, semi-chibi (not as short as full chibi) distorted character art style.

Maintained as A: The style positioning is clear, trigger words are complete, and the model has long-term iteration. However, the MOSS version lacks parameters, and changes in this version can only be seen by navigating to the fine-tuned model page.

Images

Reasons for Recommendation

• Version update: The registered trigger words for this version are only "fwaun style" and "semi-chibi style"; the previous MOSS v5.0 included "photorealistic background," but the author does not specify if this switch remains in this version.
• The LoRA itself has no update notes; the author states in the model card that details for each version are found in the corresponding Fwaun Tune fine-tuned model. The Fwaun Tune MOSS v5.1 description notes that overfitting in the previous version caused poor LoRA results, so the learning rate was lowered; this LoRA was extracted from the difference between the fine-tuned model and the base model.
• Serves those wanting flat-color, line-art-free character illustrations with smaller head-to-body ratios in MOSS; the series already has Anima, Krea 2, and Qwen 2.1 versions. The author recommends using the Fwaun Tune fine-tuned model directly unless there is a specific reason not to.
• Free; 3 visible examples; 51 downloads for this version, with 3782 cumulative downloads for the entire model (indicating attention only).

How to Use

• Load fwaun-style_moss-image_v51.safetensors (543 MB) and use "fwaun style" and "semi-chibi style" in the prompt.
• Strength, steps, and CFG were not provided by the author; older versions (Anima, Krea 2) notes suggest using "chibi, super deformed" to lower the head-to-body ratio and "realistic" to raise it, but the author does not specify if the MOSS version follows this.
• When the full model is usable, the author suggests switching to the MOSS version of Fwaun Tune.

Risks and Limitations

• The author did not write update notes for the LoRA version itself; changes can only be inferred from the description of the same-named version of Fwaun Tune.
• The LoRA is a lossy difference extraction; the author admits its effect is inferior to the fine-tuned model.
• The license does not allow selling; this entry has not been run.

Trigger Words: fwaun style, semi-chibi style

【A】Full-Bleed Video Cover — Edge-to-Edge Photo Montage, Big Title, Rental-Shop Typography | MOSS · v1.0

LoRA · New Model · Author 858654128784 · Published 11-02 19:22 · Downloads 22

90s-era video rental store cover style LoRA: photos fill the entire edge-to-edge frame, featuring a collage of characters and props, one oversized title, with blank spaces filled with colorful taglines, starburst stickers, vertical actor names, star ratings, and small print synopses. Ideal for posters, covers, and parody movie titles.

Maintaining A: The new capabilities are clear, and the prompt templates and text guidelines are actionable, but strength and sampling parameters are missing. Only 2 visible examples are provided, which is insufficient for an S rating.

Images

Reasons for Recommendation

• New model: Turns the entire layout of a "full-bleed cover on a video rental store shelf" into a single LoRA. The author claims it works with any subject matter, including village basketball finals, last-train mysteries, burger joint comedies, firefighters, construction night shifts, and gas stations.
• Serves text-to-image tasks involving large typography, such as posters, video covers, and fake movie covers; the author provides a prompt writing method: first write style keywords, then describe the genre, how characters and props are arranged, color scheme, and title placement.
• Clear warnings regarding text: Titles and taglines must be written verbatim in quotes. Use only English or Chinese (the author notes Chinese is suitable for large vertical text); Japanese will result in garbled characters. For fixed "Elegant" and "HK-style" layouts, it is recommended to use MOSS's layout mode (JSON with a box for each element).
• Free, fully open license; published approximately 2 hours ago with 22 downloads (indicating attention level only).

How to Use

• Load fullbleedvideocover.safetensors (142 MB); the author did not specify strength. No trigger word is registered, but the author requires prompts to start with "fullbleedvideocover, full-bleed video cover, photographic montage, prominent English title".
• Next, write the genre, collage arrangement of characters and props, color scheme, and title placement; write titles and taglines verbatim in quotes, using only English or Chinese.
• For fixed layouts, switch to MOSS layout mode and write JSON with boxed positions for each element; steps, CFG, and sampler were not provided by the author.

Risks and Limitations

• Strength, steps, and CFG are not provided; only prompt writing guidelines are given.
• Only 2 visible examples are available, with another 18 sensitive-level examples hidden. The sample size is too small to determine cross-genre stability.
• The author states Japanese text will be garbled; this entry has not been tested.

Worth Watching (10)

【B】AU D_elete - Artist Style · MOSS-Image

LoRA · Ported · Author Azriaelis · Published 11-02 20:16 · Downloads 7

MOSS version of the LoRA for artist AU D_elete's style: The author describes it as a big-eyed, hand-drawn style. Tags include sketching, impasto, ink lines, and long eyelashes.

Images

【B】AG: Astaroth - Vintage Anime · MOSS-Image

LoRA · Ported · Author Azriaelis · Published 11-02 20:27 · Downloads 6

MOSS version of the "Astaroth" vintage anime style LoRA from the "Ars Goetia" series; the author provided no textual description of the style, but tags suggest old-school anime and detailed digital illustration.

Images

【B】Love and Pop Lora · Moss-Image v1.0

LoRA · New Model · Author JBr4t · Published 11-02 11:14 · Downloads 111

MOSS LoRA replicating the photographic style of the film Love & Pop: The author states the focus is on the film's highly realistic visual texture, as well as its distinctive camera angles and lens effects.

Images

【B】Moss-image - MilkteaSfisdis | Style · v1.0

LoRA · New Model · Author GreatWhiteWolf · Published 11-02 18:49 · Downloads 10

LoRA for artist MilkteaSfisdis's style: The author provided no textual description of the style, but tags suggest BL-oriented illustrations featuring primarily adult male characters.

Images

【B】V67 Vintage Pulp Art · V1 MOSS

LoRA · Ported · Author angelomaiota · Published 11-02 19:01 · Downloads 13

MOSS version of the LoRA for 1950s American pulp magazine cover art style: 2D flat coloring, thick outlines, and halftone print texture. Suitable for retro cover illustrations in sci-fi adventure, detective, and monster genres.

Images

【B】moss marionette lines · v1

LoRA · New Model · Author hamstornado · Published 11-02 16:44 · Downloads 25

Marionette lines (mouth corner sagging) LoRA: Adds slight shadowing and drooping to the corners of the mouth, making characters appear sullen, cold-faced, or "grumpy".

Images

【B】Bojack Horseman Style · MOSS

LoRA · Ported · Author OrdinaryTwig · Published 11-02 13:56 · Downloads 24

MOSS version of the LoRA for the animated series BoJack Horseman's style: Produces images with the show's flat cartoon aesthetic and anthropomorphic animal characters. The author claims most characters can be generated simply by describing their appearance.

Images

【B】White Cat Legend - Li Bing · v1

LoRA · New Model · Author 1000Mbps · Published 11-02 12:23 · Downloads 14

Character LoRA for Li Bing, the White Cat Minister from the domestic animated series White Cat Legend. Includes a separate trigger word to generate his official uniform from the original work.

Images

【B】Eagle Marin | Eagle Seat Marin Training Outfit | Saint Seiya · v1.6 Moss

LoRA · Ported · Author w63725310859 · Published 11-02 10:49 · Downloads 20

MOSS version of the character LoRA for Marin of the Eagle (training outfit) from Saint Seiya. The author includes a set of English natural language prompts broken down by body part to accurately reproduce the character.

Images

【B】Mecha Engrave · moss-image-mecha-v1

LoRA · Ported · Author Xabsurd · Published 11-02 17:11 · Downloads 0 · Paid

MOSS version of the Mecha Detail Enhancement LoRA: Thickens armor layers, enhances the three-dimensionality of decorations like beast heads and spikes, and boosts glowing energy parts, making mecha images heavier and more complex.

Images

Included Only (15)

• AG: Aim - Synthwave Surrealistic Style · MOSS-Image — LoRA: The MOSS version of the "Aim" style LoRA from the "Ars Goetia" series; the author provided no textual description of the style, but based on the tags, it features a synthwave color palette and a glitchy, surreal poster aesthetic.
• Xu Qinghuan · v1.0 — LoRA: A character LoRA for "Xu Qinghuan," an AI-original female character created by the author, used to maintain a consistent face and persona.
• Ryu Hayabusa · Ryu_Hayabusa_MOSS_v1.0 — LoRA: The MOSS version of the character LoRA for Ryu Hayabusa, the protagonist of Ninja Gaiden: featuring his iconic look with a silver winged helmet, black mask, and dark blue ninja outfit, it can also generate a short-haired version without the mask.
• alterkyon style lora | Moss · v1.0 — LoRA: The MOSS version of the style LoRA for artist alterkyon; the author did not describe the style and admitted the training was not fully successful and still has issues.
• Genshin Impact Miliastra Wonderland-Chromatic Fatebound Sea | Qianxing Qiyu-Liucai Yuanhai · v1.0 MOSS — LoRA: The MOSS version of the outfit LoRA for the "Chromatic Fatebound Sea" set from Genshin Impact's Miliastra Wonderland: featuring a bikini, leg ring with a clover emblem, and sarong, with optional star-decorated straw hat, Roman sandals, and water gun.
• Isobelle NordhavenLove · MossImage — LoRA: A character LoRA for the fictional character "Isobelle NordhavenLove": the author defines her as a knitwear designer living by the Swedish coast, often wearing hand-knitted sweaters and linen pants.
• Rey - Star Wars Comics (Sith & Jedi) · MOSS v1.0 — LoRA: The MOSS version of the character LoRA for Rey from the Star Wars comics, featuring two distinct looks: Jedi (brown hair, white crop top) and Sith (gray skin, yellow-red eyes, black lips, black dress).
• Pacifica Northwest - Scrag Boy · Moss — LoRA: The MOSS version of the "grown-up" character LoRA for Pacifica Northwest from Gravity Falls, featuring a fixed outfit with a purple bodysuit, open jacket, and fringed boots, including 3D/CGI style keywords.
• Mr Piglets Oldies and brats Moss image · Lux46 — LoRA: A specific character version from the character collection LoRA "Mr Piglets Oldies and brats": each version represents a different realistic female character, with Lux46 being an older woman with short platinum blonde hair.
• indian-girl-brown-eyes-moss-sft-2 · moss-sft — LoRA: A realistic portrait LoRA for an Indian female model in her twenties, with the trigger word indian-girl-brown-eyes, used for generating portraits with a consistent face.
• Luo Tianyi V3 cosplay luo tianyi · v1.0 — LoRA: A cosplay-style LoRA for Luo Tianyi's V3 look: generates images of characters wearing Luo Tianyi's V3 outfit; the author states that all training images were generated by banana2, based on MOSS SFT.
• clorinde · v1 — LoRA: An anime-style LoRA for the Genshin Impact character Clorinde, with the author only writing "just for share".
• chilin_majsoul · v1 — LoRA: A LoRA for the Mahjong Soul character Chilin, with the author only writing "just for share" and the trigger word being the Chinese term "赤麟" (Chilin).
• xAngel · MOSS Image — LoRA: A LoRA for the fictional realistic female character "xAngel": the author describes her as a young blonde woman with a slim, athletic build, with all training images being AI-generated.
• 3D | Moss · v1.0 — LoRA: A style LoRA that the author claims adds a "3D effect" to images, but the description is only one sentence, and what 3D specifically refers to is not explained.

---

Data: Civitai Public API (base model MOSS Image), collection window 11-02 10:15 → 11-02 21:01 (Beijing Time). Recommendations have been reviewed by editors; download counts only reflect popularity.