Skip to main content
NSFW LoRA guide

Best MOSS-Image NSFW LoRAs — what to load and how

MOSS-Image is OpenMOSS’s open 5-billion-parameter diffusion transformer, released under Apache-2.0 with a Base and an SFT checkpoint. On its own it delivered every level of our content test as written, so on this model a LoRA is for the look — realism, portrait or style — not for permission. Here are the community files we ran, a same-seed comparison, and how to load them.

Get an API keyOpen MOSS-Image LoRA
From
$0.03/ image
Text to image · 1K
LoRAs compared
8
Tested on SpicyAPI
6

What we measured on MOSS-Image

  • Base and SFT files both load

    LoRAs trained on the Base and on the SFT checkpoint both loaded, in the diffusers (pytorch_lora_weights.safetensors) and the ComfyUI file formats, and three stacked in one request worked. 21 Oct 2026.

  • Civitai sign-in files: HTTP 401

    Realistic Snapshot — 100,520 downloads, the most-downloaded MOSS-Image LoRA page — and DaSkin both failed without a token on 22 Oct 2026, and the failed jobs weren’t charged. Download them with your own account and host them at a direct link.

  • Smartphone Snapshot: 15.9% change

    Same prompt, same seed: Smartphone Snapshot Photo Reality at 1.0 changed the image by 15.9%, Moss Image Finish Tweaker at 1.0 by 4.1%. Repeat runs with the same seed return the same image, so the numbers are the LoRAs’ own effect. 22 Oct 2026.

Tests ran on each family's main route with the exact request shape the API sends — SpicyAPI passes LoRA links through unchanged. Outputs stay in our test archive and are not published.

Same prompt, same seed, three LoRA choices

One prompt — an adult woman laughing at a rainy café window with a coffee — seed 7, 1k, on MOSS-Image LoRA, 22 Oct 2026. Only the LoRA changes; repeat runs with this seed return the same image.

  • An adult woman laughing at a rainy café window with a cup of coffee, generated by MOSS-Image without a LoRA

    No LoRA

    MOSS-Image on its own.

  • The same café scene as a flatter, cooler phone snapshot with wider framing, from the Smartphone Snapshot LoRA

    Smartphone Snapshot Photo Reality

    At 1.0: a flatter, cooler phone-snapshot look and a wider frame.

    15.9% different from no LoRA

  • The same café scene with the same composition and slightly different light and texture, from the Finish Tweaker LoRA

    Moss Image Finish Tweaker

    At 1.0: the same composition, with slightly different light and texture.

    4.1% different from no LoRA

The MOSS-Image NSFW LoRAs worth loading

Public download counts from Oct 10, 2026, the strength each author recommends, what we used, and whether the file runs here.

Tested
we ran this exact file
Link blocked
the repo now requires a sign-in

Use LoRAs trained for MOSS-Image, on the Base or the SFT checkpoint: both loaded in our tests, in the diffusers and the ComfyUI file formats. Put up to three in loras on MOSS-Image LoRA, strength 0 to 4, and link the .safetensors file directly (on Hugging Face, the /resolve/main/ address). Counts in this group are from 22 October 2026.

Best NSFW LoRAs for MOSS-Image
LoRATypeDownloadsStrengthFileLicenseOn SpicyAPI
MOSS-Image NSFW UncensoredaxiomofmindNSFW all-in-one9.4K256 likes0.7 author0.7 our test74 MBAny serviceTestedOct 21, 2026
Realistic SnapshotMonkeyForeverRealism100.5K4.1K likes—567 MBAny serviceLink blocked
Smartphone Snapshot Photo RealityAI_CharactersRealism78843 likes1.0 our test39 MBAny serviceTestedOct 22, 2026
MOSS-Image BeautyaxiomofmindRealism4556 likes0.7–1.0 author1.0 our test74 MBAny serviceTestedOct 21, 2026
Moss Image Finish Tweakerlse14Realism28025 likes1.0 our test9 MBAny serviceTestedOct 22, 2026
DaSkin Realistic SkinStarnodesRealism9919 likes—126 MBAny serviceLink blocked
MOSS-Image GouacheaxiomofmindStyle2831 like1.0 · 0.6 our test74 MBAny serviceTestedOct 21, 2026
MOSS-Image Psychedelic PosteraxiomofmindStyle1361 like1.0 our test74 MBAny serviceTestedOct 21, 2026

Counts come from the public Civitai and Hugging Face APIs on Oct 10, 2026; a Civitai count can combine several versions of one LoRA. LoRAs whose names spell out specific acts or body parts — most of Wan 2.2's top 50 — are left off by name, but they load the same way. Repositories built around real people, or anyone who looks underage, are excluded entirely.

Load a LoRA on MOSS-Image

MOSS-Image LoRA is text-to-image only: one loras slot with up to three path + scale entries, scale 0 to 4 (default 1). Link the .safetensors file directly — on Hugging Face, the /resolve/main/ address; a webpage link fails and isn’t charged. There is no steps or guidance setting, so Turbo and 4-step speed LoRAs bring no speed-up. File formats and failed links are covered in our MOSS-Image LoRA guide, listed below.

Hugging Face, Civitai and private link formats
Example request
POST /api/v1/jobs/createTask
Authorization: Bearer $SPICY_API_KEY
Content-Type: application/json

{
  "model": "openmoss/moss-image-lora/text-to-image",
  "input": {
    "prompt": "<trigger word>, candid photo of two adults on a hotel balcony at dusk",
    "loras": [
      {
        "path": "https://civitai.com/api/download/models/<versionId>",
        "scale": 0.8
      },
      {
        "path": "https://huggingface.co/<owner>/<repo>/resolve/main/<file>.safetensors",
        "scale": 0.6
      }
    ]
  }
}

Tips for MOSS-Image NSFW LoRAs

  • Put the trigger word first

    Start the prompt with the LoRA’s trigger: moss_xvstyle for NSFW Uncensored, moss_beauty_style, moss_vgouache_style or moss_psychpop_style; Smartphone Snapshot uses photo.

  • Start at the author’s strength

    0.7 for NSFW Uncensored, 1.0 for Beauty (0.7–0.9 works too). Stay between 0.6 and 1 before going higher: the slot accepts up to 4, but that’s a ceiling, not a target.

  • Copy the MOSS-Image file

    Some Civitai pages carry several bases: Realistic Snapshot’s page also has Krea 2 and Z-Image Turbo versions. Only the MOSS-Image version works here — copy that version’s download link.

Where to run MOSS-Image LoRAs on SpicyAPI

Guides and tests for these models

MOSS-Image NSFW LoRA questions

NSFW LoRA guides for other models

All LoRA models, the full test log and how-to