Z-Image Turbo LoRA
Text to image · Image editing
- Train your LoRA on
- Z-Image Turbo
- Freedom
- 95 / 100
- From
- $0.012 / imageImage editing
Z-Image Turbo's text-to-image already renders every level of our content test without help, which makes it the cheapest place to run NSFW LoRAs for the look rather than for permission: realism, anatomy and style. Here's what the community downloads and what happened when we ran it.
NSFW MASTER loaded at 1.24 GB
The Hugging Face port of Shopon_Skp's NSFW MASTER is a 1.24 GB file — over the documented 1 GB limit — and it still loaded and rendered L4. 26 Sep 2026.
Civitai links: loaded, even sign-in-only
Mystic v7 (open) and NSFW MASTER V2 (sign-in only) both loaded straight from Civitai download links with no token, and a civitai.red link returned a pixel-identical image. Each changed the image about 20%, against 2.6% for a repeat run. 28 Sep 2026.
Edit without a LoRA: L4 implied 3 of 3
Z-Image Turbo's image-to-image kept two-person edits implied in all three runs — that's where an NSFW LoRA earns its place. 27 Sep 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.
Public download counts from Oct 10, 2026, the strength each author recommends, what we used, and whether the file runs here.
Z-Image Turbo already reaches L4 without help, so its LoRAs are mostly about realism, anatomy and style. It's a distilled model: keep the combined strength of stacked LoRAs around 1.
| LoRA | Type | Downloads | Strength | File | License | On SpicyAPI |
|---|---|---|---|---|---|---|
| NSFW MASTER (Z-Image Turbo)Shopon_Skp | NSFW all-in-one | 98.1K6.5K likes | 0.8 author1.0 · 0.8 our test | 593 MB | Any service | TestedSep 28, 2026 |
| Mystic (ZIT)alcaitiff | NSFW all-in-one | 114K3.7K likes | 0.8 our test | 198 MB | Any service | TestedSep 28, 2026 |
| Z-Image Turbo NSFW LoRAPeli86 | NSFW all-in-one | 29.5K819 likes | — | 13 MB | Civitai only | Should run |
| Realistic SnapshotMonkeyForever | Realism | 92.3K3.9K likes | 0.6–0.7 author | 162 MB | Any service | Sign-in file |
| Lenovo UltraRealDanrisi | Realism | 81.9K13 likes | 1.2–2 author | 81 MB | Any service | Should run |
| Pixel Art Styletarn59 | Style | 940.6K56 likes | 1.0 our test | 162 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.
One loras slot, up to three files, on both text-to-image and image-to-image. Z-Image Turbo is a distilled model, so strengths add up fast: keep the combined total around 1.
Hugging Face, Civitai and private link formatsPOST /api/v1/jobs/createTask
Authorization: Bearer $SPICY_API_KEY
Content-Type: application/json
{
"model": "alibaba/z-image-turbo-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
}
]
}
}Three LoRAs at 1.0 each over-expose a turbo model. A realism LoRA at 0.6 plus an NSFW pack at 0.5 is a better start than either at full strength.
Lenovo UltraReal's author asks for 1.2–2 on Turbo; the latest Realistic Snapshot says 0.6–0.7. Recommended strengths vary two- to threefold between LoRAs and versions.
Many Civitai pages carry FLUX, Krea 2 and Z-Image Turbo versions of the same LoRA. Only the Z-Image Turbo file works here — copy that version's download link.
Text to image · Image editing

Image24 min read

Image9 min read

Image10 min read

Image33 min read

Image15 min read