Zhenzhen's AI Radar · MOSS Image · 2026-11-11 Evening Report
This issue features 15 new versions: Strongly Recommended 2 · Recommended 2 · Worth Watching 1 · Included Only 10
Editor's Note This issue reviewed 18 new MOSS versions, with 2 rated S, 3 rated A, 1 rated B, and 11 rated C; 1 were excluded. The top pick is HaDeS's MOSS Turbo single-file base model: MOSS Base has three LoRAs baked in, allowing image generation in 8 steps with CFG 1 without needing an additional acceleration LoRA; parameters and tuning observations are fully documented. The corrugated cardboard collage style LoRA is also rated S, while square-toe platform shoes, Brabus models, and Chinese period film still styles are rated A. Three entries rated C include NFFA (self-assessed as average), standard character LoRAs, and the MOSS LoRA Lab (paid with no documentation). A chest LoRA with a name implying minors has been excluded.
Strongly Recommended (2)
[S] HaDeS MOSS Turbo · v0.9.4 BF16
Base Model / Checkpoint · New Model · Author _HaDeS_ · Released 11-11 20:32 · Downloads 10
HaDeS's MOSS single-file 8-step text-to-image base model: Based on MOSS-Image-Base, with three LoRAs baked in (including the author's unpublished Turbo LoRA), it generates images in 8 steps with CFG 1 without requiring an additional acceleration LoRA.
The HF model card includes paired comparisons and scoring records against SFT + Turbo; check them out if you want to see the differences first.
Images

Reasons for Recommendation
- New model; serves users who want fast text-to-image generation with MOSS but don't want to stack Turbo / detail LoRAs themselves, offering an alternative output style to the official SFT; the author explicitly states that comparison results do not imply it is better than SFT for all prompts
- Parameters are fully provided: 8 steps, CFG 1.0, Euler + Simple, flow shift 5.0 (set to 0 in HaDeS MOSS Suite), denoise 1.0, SLG off, tested on ComfyUI 0.43.0; comparison images use 1152 × 1536
- The author also documented tuning observations: at 8 steps, increasing CFG from 1 to 2 results in more saturation, darker images, and approximately 1.8 times the processing time; increasing steps changes composition but does not consistently improve quality; 17 visible example images
How to Use
- Place the file in ComfyUI/models/diffusion_models/ and use a MOSS-compatible loader (the author uses the MOSS Loader from HaDeS MOSS Suite)
- Text encoder qwen3.5_9b_int8_convrot and VAE flux2-vae are both available from Comfy-Org/MOSS-Image; do not connect a separate turbo LoRA input to the KSampler
- 8 steps, CFG 1.0, Euler, Simple, flow shift 5.0; when using the Suite, set MOSS Prompt Enhancer to 'as written' or 'caption'
Risks and Limitations
- Contains only the diffusion model; the text encoder and VAE must be downloaded separately; the baked-in Turbo LoRA is not public, so the recipe cannot be reproduced independently
- The Hugging Face model card lists Apache-2.0, while the Civitai license only checks image generation and site rental; adhere to the stricter of the two; 3 examples are hidden
[S] Dizzle's Corrugated Cardboard - MOSS · v1
LoRA · New Model · Author Maridizzle · Released 11-11 20:49 · Downloads 12
Corrugated cardboard / cardstock collage style LoRA: Turns any scene into layered paper art models, ranging from monochrome kraft paper reliefs to high-saturation colored cardstock.
Images

Reasons for Recommendation
- New model; serves the "handmade paper art" effect in illustrations, posters, and still life scenes: contour-like paper layers, sculpted edges, and cardboard fuzz textures
- Usage is fully documented: Place the trigger word dz_card at the beginning of the prompt, paired with 'layered cardboard cutout', and describe the scene in natural language; strength 0.9–1.0; MOSS default settings are approximately 30–50 steps, CFG 3, Euler flowmatch
- The author provides generation tips: Scenes with distinct silhouettes and clear depth work best; specify colors in the prompt for colored results, or use only warm brown tones for kraft paper reliefs; 4 visible example images
How to Use
- Start the prompt with 'dz_card, layered cardboard cutout', then describe the scene in complete sentences
- Strength 0.9–1.0; 30–50 steps, CFG 3, Euler flowmatch
Risks and Limitations
- Only 4 visible examples, with another 4 hidden; the license does not include selling rights
- The step count of 30–50 is relatively high, and the author did not specify performance when paired with acceleration LoRAs
Trigger Word: dz_card
Recommended (2)
[A] Brabus P 900 Rocket Edition · Moss
LoRA · New Model · Author denrakeiw · Released 11-11 14:42 · Downloads 18
A vehicle model LoRA for the 2026 Brabus P 900 Rocket Edition (a wide-body pickup based on the Mercedes AMG G 63), capable of generating exterior angles, cargo bed, engine bay, and interior.
Images

Reasons for Recommendation
- New model; serves scenarios in automotive posters and advertising concept art requiring accurate reproduction of this limited-edition model: details such as black paint, carbon fiber wide-body wheel arches, 24-inch three-spoke wheels, and red-accented interiors are within the training scope
- The author states training used 69 official news and studio photos; the trigger phrase is 'Brabus P 900 Rocket Edition', with strength 0.8–1.0; prompts follow the training format starting with 'Create a photograph of ...', with body badge text enclosed in double quotes
- Provides example prompts written by viewpoint; 14 visible example images
How to Use
- Start the prompt with 'Create a photograph of the Brabus P 900 Rocket Edition', specifying camera angle, lighting, and materials; place badge text in double quotes
- Strength 0.8–1.0; steps and CFG were not provided by the author
Risks and Limitations
- Training images are official brand photos; the vehicle model and logos involve trademarks, so verify commercial use rights independently
- The model card title says Flux LoRA, which appears copied from the Flux version description; differences in the MOSS version are not specified; 3 examples are hidden
[A] Ancient Cinema Classical Film Concept | MOSS · MOSS
LoRA · New Model · Author 858654128784 · Released 11-11 16:32 · Downloads 51
Chinese period film still style LoRA: Lakeside ferries, market streets, mountain paths, temples, city walls, and lantern-lit nights, with blurred willow branches or reeds in the foreground and distant mountains and palaces fading into mist, featuring restrained colors and a soft film-like feel.
Images

Reasons for Recommendation
- New model; serves period drama stills, ancient-style posters, and storyboard concept art; the author refers to this style as "Jin Ping Mei / Golden Lotus case style"
- The prompt structure is specific: Start with 'ancientcinemaconcept, A cinematic historical photograph.', describe characters and scenes, add a lighting sentence (e.g., 'Blue hour slate blue ambient light'), and end with a fixed foreground blur and layered depth sentence
- Can be stacked with lightshadowart (strength 0.6) to create a beam of window light or moonlight; the author provided stacking examples; 6 visible example images
How to Use
- Prompt: 'ancientcinemaconcept, A cinematic historical photograph.' + Character/Scene + Lighting Sentence + 'Foreground softness, selective focus and subdued distant details create layered photographic depth.'
- Strength, steps, and CFG for this LoRA were not provided by the author; use 0.6 when stacking lightshadowart
Risks and Limitations
- 10 examples are hidden, and the page contains mixed content
- The description is short and does not specify the training data source
Worth Watching (1)
[B] Scavengers Reign · moss
LoRA · New Model · Author fyrtyv · Released 11-11 15:28 · Downloads 21
A MOSS Base style LoRA inspired by the animated series "Scavengers Reign": Thin lines, low-saturation colors, and intricate organic forms, focusing on alien landscapes and exotic ecosystems.
Downgraded from A to B automatically: The purpose is clear, but there are no usage parameters other than the trigger word.
Images

Reasons for Recommendation
- New model; serves sci-fi concept art: Scenes like towering forests, glowing caves, and abandoned spaceships reclaimed by creatures
- Trigger word: vestaline; 4 visible example images, none hidden
How to Use
- Add 'vestaline' to the prompt and describe alien landscapes or ecological scenes; strength, steps, and CFG were not provided by the author
Risks and Limitations
- The description is only a single paragraph of style description, with no parameters
- Mimics the art style of a specific animated series; be cautious about commercial use; the license does not include commercial image generation
Trigger Word: vestaline
Included Only (10)
- RLY Thot Shot - Adeline (MOSS + KREA 2 + ZiB/ZiT) · MOSS-v1.0 — LoRA: The MOSS version of the character LoRA for Adeline, the 40th fictional female character in the RLY Thot Shots series (Berlin librarian setting).

- NFFA Style | MOSS Image LoRA · v1.0 — LoRA: A style LoRA named NFFA. The author self-assesses the results as average and provides three files trained for different numbers of epochs for user selection.

- Frankie Von Vergin · MossImage — LoRA: The MOSS version of the character LoRA for Frankie Von Vergin, a fictional character from the MisterMR series (Oxford rare map librarian setting).

- xFiona · v1 — LoRA: The MOSS character LoRA for xFiona, an AI-generated fictional female character by randomimagepotato.

- Event Horizon | MossLoraLab · DarkFantasy v1.0 — LoRA: A LoRA named DarkFantasy from the MOSS LoRA Lab collection. The author did not describe the specific effects.

- Emilia/艾米利亞 (Re:Zero) Moss sft lora v1.0 · Emilia/艾米利亞 Moss sft v1.0 — LoRA: The MOSS SFT character LoRA for Emilia from Re:Zero. The description only contains the title.

- Aliyah Belwyn · MossImage — LoRA: The MOSS version of the character LoRA for Aliyah Belwyn, a fictional character from the MisterMR series (Bruges harpist setting).

- Minnie Belwyn · MossImage — LoRA: The MOSS version of the character LoRA for Minnie Belwyn, a fictional character from the MisterMR series (Bruges miniature portrait painter setting).

- Event Horizon | MossLoraLab · Neon Dinasty v1.0 — LoRA: A LoRA named Neon Dinasty from the MOSS LoRA Lab collection. The author did not describe the specific effects.

- Event Horizon | MossLoraLab · Hyperdetailed v1.0 — LoRA: A LoRA named Hyperdetailed from the MOSS LoRA Lab collection. The author did not describe the specific effects.

Data: Civitai public API (base model MOSS Image). Inclusion window: 11-11 12:33 → 11-11 21:33 (Beijing Time). Recommendation reasons have been reviewed by editors; download counts only reflect popularity.