Qwen Image 2512 LoRA
Text to image · Image editing
- Train your LoRA on
- Qwen-Image 2512 · Edit 2511
- Freedom
- 92.5 / 100
- From
- $0.03 / imageImage editing
Qwen Image is where NSFW editing LoRAs live: the most-downloaded one has passed 500,000 downloads on Hugging Face. This page covers the 20B generation — 2512 and Edit 2511. The 7B Qwen Image 2.1 takes different LoRAs and has its own page.
Edit + MCNL: L4, same person
The Multi-Concept NSFW LoRA on the Edit 2511 endpoint turned a clothed reference into an L4 scene and kept the person from the reference. 26 Sep 2026.
No LoRA: L4 implied 3 of 3
Qwen Image 2512 text-to-image and edit both kept two-person scenes implied in all three runs, while L1–L3 rendered every time. That's the gap its NSFW LoRAs close. 27 Sep 2026.
Links go stale
The anatomy-fix LoRA we ran on 26 September was taken offline two days later; its author re-published a v2 under a non-commercial license. Re-check links you depend on.
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.
Public download counts from Oct 10, 2026, the strength each author recommends, what we used, and whether the file runs here.
One 20B architecture across 2512 and Edit 2511, with the richest NSFW editing LoRAs. Without a LoRA, two-person scenes stayed implied in 3 of 3 of our runs. Many packs here are 1.6–2.25 GB — check the size before you paste.
| LoRA | Type | Downloads | Strength | File | License | On SpicyAPI |
|---|---|---|---|---|---|---|
| MCNL Multi-Concept NSFW (edit)ScottzillaSystems | NSFW all-in-one | 525.8K536 likes | 1.0 our test | 563 MB | Any service | TestedSep 26, 2026 |
| SNOFSAshen3 | NSFW all-in-one | 289.2K7.5K likes | — | 585 MB | Civitai only | Sign-in file |
| QWEN 4 PLAY AIO NSFW (2512)ERA5ER | NSFW all-in-one | 49.1K1.6K likes | — | 563 MB | Any service | Sign-in file |
| Qwen-Image NSFW Advfok3827 | NSFW all-in-one | 28.5K1.1K likes | 1.0 author | 840 MB | Any service | Should run |
| Female Anatomy Fix v2perpetual3x | Anatomy fix | — | 0.4 author | 563 MB | Non-commercial | Should run |
| Anything2Real (edit)lrzjason | Realism | 1.2M143 likes | 0.85 our test | 585 MB | Any service | TestedSep 26, 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.
Each family takes one loras slot with up to three files. Use 20B files (Qwen-Image, 2512, Edit 2509 and 2511) on Qwen Image 2512 LoRA — its edit endpoint runs Edit 2511 — and files marked Qwen 2.1 on Qwen Image 2.1 LoRA.
Hugging Face, Civitai and private link formatsPOST /api/v1/jobs/createTask
Authorization: Bearer $SPICY_API_KEY
Content-Type: application/json
{
"model": "alibaba/qwen-image-2512-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
}
]
}
}Qwen's 20B LoRAs are big: 563–840 MB is typical and many NSFW packs are 1.6–2.25 GB, over the limit. Look for a smaller rank on the same page.
The Qwen anatomy-fix author recommends 0.4; NSFW packs sit at 0.8–1.0. Drop them to 0.6–0.8 when you stack a character LoRA.
Camera-angle, illustration-to-photo and relighting LoRAs change what the edit endpoint does, and they combine with an NSFW pack in the same request.
Text to image · Image editing

Image9 min read

Image24 min read

Image27 min read

Image18 min read

Image22 min read