Is this the same model as Wan 2.2 LoRA?

No. Wan 2.2 LoRA runs the 14B Mixture-of-Experts models, with separate high-noise and low-noise LoRA slots. This family runs the dense 5B TI2V model, which has one LoRA slot and a different network shape, so LoRAs do not carry over between the two. In our same-seed test a 14B LoRA on the 5B model left the clip unchanged.

Which LoRAs work here?

LoRAs trained for Wan2.2-TI2V-5B: on Hugging Face they list it as their base model. In our tests a community fabric-style LoRA, a vintage-colour LoRA and a community unfiltered LoRA all loaded and changed the result.

Why did my LoRA fail?

The two causes we saw: the Hugging Face file needs a login (the error mentions HTTP 401), or the link points to the repository page instead of the .safetensors file. Neither is charged.

My LoRA loaded but nothing changed. Why?

It was most likely trained for a different base model, such as the 14B Wan 2.2 or Wan 2.1. Those files load without an error and do nothing here. Some LoRAs also need their trigger word in the prompt.

Can I make longer clips?

No. Every clip is 121 frames at 24 fps, about five seconds. For longer clips use a larger Wan model.

Is Wan 2.2 5B LoRA uncensored for NSFW video?

Yes. Your prompts, LoRAs and results are not reviewed or filtered. In text-to-video the model rendered NSFW scenes with women as written, from Mild through Strong and at Maximum; it held back at the Very strong level and on the strongest requests with a male subject. Image-to-video went as far as the Moderate level. Every person depicted must be an adult.

Per clip, at one flat price on both endpoints. Loading LoRAs adds no separate charge, and failed requests are not charged.

Wan 2.2 5B LoRA API: Uncensored NSFW 720p Video, Your LoRAs

alibaba/wan-2.2-5b-lora/text-to-video

Wan 2.2 5B LoRA text-to-video turns a prompt into one five-second 720p clip on Alibaba's TI2V-5B model, with up to three of your own LoRA files. Landscape or portrait, one flat price per clip, no content filter.

From $0.12/request

Run model View API

Input

Text to video

Prompt

Aspect ratio

Style LoRAs0/3

Seed

Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Example result

Extend

Video upscaling

Video editing

{
  "model": "alibaba/wan-2.2-5b-lora/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "seed": 50101,
    "prompt": "Cinematic video. Blue hour in the lounge of an alpine ski lodge: an adult Nigerian-British woman around thirty with close-cropped natural hair, in an ivory off-the-shoulder knit dress, is curled up in a cognac leather wingback chair, snow drifting past the tall glass behind her. She lowers the book in her lap, tilts her head and gives the camera a slow, smouldering glance. Warm copper lamp light against cold blue dusk, slow push-in, 50mm lens, shallow depth of field."
  }
}

Introduction

Wan 2.2 5B LoRA Text to video turns a prompt into a five-second 720p clip with up to three of your LoRA files loaded. It runs Alibaba's open Wan2.2-TI2V-5B, a dense 5-billion-parameter model; clips come back in 16:9 or 9:16, silent, at one flat price per clip. The platform adds no content filter to the prompt, the weights or the clip.

Choose a mode by starting material

What you have	Mode
a written description	Wan 2.2 5B LoRA Text to video
a photo to animate	Wan 2.2 5B LoRA Image to video
LoRAs made for the 14B Wan 2.2	Wan 2.2 LoRA Text to video

Key features & highlights

LoRAs for the 5B base

Community style, character and unfiltered LoRAs trained on Wan2.2-TI2V-5B load by link, up to three at once.

A flat price per clip

No duration or resolution tiers: one request, one five-second 720p clip, one price.

Steady faces and fabric

Knitwear, skin and lamplight hold up across the clip; one action per prompt works best.

Uncensored with custom weights

LoRAs are not reviewed and the output is not filtered.

Request parameters

Request parameters for Wan 2.2 5B LoRA · Text to Video
Parameter	Required	Description
`prompt`	Yes	Trigger word first if your LoRA has one, then who is in frame, one action and the camera move. Five seconds fits one moment. at least 1 characters
`aspect_ratio`	No	16:9 gives 1280 x 704 (the default); 9:16 gives 704 x 1280. The shape does not change the price. `16:9` · `9:16` · Default: `16:9`
`loras`	No	Up to three {path, scale} entries trained for Wan 2.2 TI2V-5B; scale 0 to 4, 1 by default (start style LoRAs at 0.5). Use a public direct link to the .safetensors file. Default: `[]` at most 3 items
`seed`	No	Keeps runs comparable while you change one LoRA scale at a time. Two identical requests can still differ slightly, so it is not a frame-identical guarantee. `≥ 0` · `≤ 2147483647`

Pricing

$0.12 / clip

Billed per clip: every request returns one five-second clip at one flat price. The aspect ratio does not change the price, and loading LoRAs adds no separate charge.

Use cases

LoRA drafts

Try a new LoRA at two or three scales before you commit a shot to a larger model.

Vertical clips

704 x 1280 portrait clips for social feeds.

Apps without a content filter

Load the LoRA your audience expects; adults only on screen.

Explore key use-case tutorials

Prompt guide & pro tips

Check the base model

LoRAs for the 14B Wan 2.2 or Wan 2.1 load without an error and change nothing here.

Compare against no LoRA

Same prompt, same seed, loras empty: two identical requests still differ slightly, so look for a clear change in look, not a few pixels.

Use direct, public links

The link must download the .safetensors file itself; a page link or a gated file fails and is not charged.

API quick start

# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
  -d '{
  "model": "alibaba/wan-2.2-5b-lora/text-to-video",
  "input": {
    "aspect_ratio": "16:9",
    "prompt": "A slow aerial shot drifting over a misty pine forest at sunrise"
  }
}'

# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
  -H "Authorization: Bearer $SPICY_API_KEY"

See the API reference for authentication, the complete schema, polling, and error handling.

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

LORA VIDEO API

Wan 2.2 5B LoRA API at a glance

Wan 2.2's compact 5B video model with your own LoRAs: five-second 720p clips from text or a photo, no content filter.

Parameters: 5B; A dense model with a high-compression video VAE, built by Alibaba to run on one consumer GPU.
LoRAs per call: 3; Up to three weight files per clip, each with its own strength from 0 to 4.
Clip length: 5 s; Every clip is 121 frames at 24 fps; there is no length setting.
Resolution: 720p; 1280 x 704 or 704 x 1280; image-to-video follows the shape of your photo.
Billing unit: 1 clip; One flat price per clip; loading LoRAs adds nothing.
Content filter: None; Uncensored: no review of your LoRAs and no filter on the result.

Results

What you can make with Wan 2.2 5B LoRA

Style LoRAs
A 1910s colour process, on cue
A community LoRA trained on early colour photographs, at scale 1 with its trigger words first: against the same seed without it, the alpine hall comes back softer and muted, with period lace and hair, and she turns to the camera.
Text-to-video with a community early-colour-photo LoRA at scale 1, 16:9, five seconds.
Text to video
One look, five seconds
No LoRA, one moment: blue hour in a ski lodge, she looks up from her lap with a slow glance. The 5B model holds faces, knitwear and lamplight steady through the clip.
Text-to-video without a LoRA, 16:9, five seconds.
Image to video
One action from a still
A still of a skier in a sunlit boot room becomes a quiet moment: he sits back, pushes his hair back and smiles. Image-to-video keeps your photo and plays out the one action you describe.
Image-to-video from a still, without a LoRA, five seconds.

More clips made with Wan 2.2 5B LoRA

Text-to-video without a LoRA, 9:16, five seconds.
5s
Image-to-video with the early-colour-photo LoRA at scale 1, five seconds.
5s
Image-to-video from a still, without a LoRA, five seconds.
5s
Image-to-video with the woven-fabric LoRA at scale 1, portrait, five seconds.
5s
Text-to-video with a community woven-fabric LoRA at scale 1, 16:9, five seconds.
5s
Text-to-video without a LoRA, 16:9, five seconds.
5s

Content test

Is Wan 2.2 5B LoRA uncensored? What we tested

We ran each level on our live setup and checked every clip frame by frame. A pass means the content was actually on screen, not just that the request went through.

Content levels tested on Wan 2.2 5B LoRA, by mode
Level	Text to video	Image to video
MildThe lightest prompts in the test	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
ModerateWhere filtered models start to refuse	RendersTested Oct 9, 2026	RendersTested Oct 9, 2026
StrongThe middle of the scale	RendersTested Oct 9, 2026	Not tested yet
Very strongWhere most models start to hold back	Not tested yet	Not tested yet
MaximumThe top of the scale	RendersTested Oct 9, 2026	Not tested yet

Renders: on screen as written. Softened: generated, but milder than what was written. Refused: rejected. Not tested yet: we have not run that level on that mode.

How each mode behaves

All modes: Every clip is five seconds at 720p and silent: there is no duration, resolution, negative prompt or audio setting. LoRAs must be trained for Wan 2.2 TI2V-5B; files for the 14B Wan 2.2 models load without an error and do nothing. A link that needs a login or points to a repository page fails the request, which is not charged. Three LoRAs per request is a hard cap.
Text to video: Only two shapes: 16:9 and 9:16. In our content tests the model delivered scenes with one woman as written, but held back at the Very strong level and toned down a request with a male subject.
Image to video: The clip keeps the shape of your photo. In our content tests it went as far as the Moderate level and no further, and held back at the Very strong level.

Use cases

What to build with the Wan 2.2 5B LoRA API

Cheap LoRA drafts
A small model and a flat per-clip price make it practical to try a new LoRA or a prompt idea many times before moving the shot to a larger model.
Portrait clips for social feeds
Generate 9:16 clips at 704 x 1280 for vertical feeds, with one LoRA keeping the look of a series consistent.
Bring a still to life
Give a portrait or a scene a quiet five-second moment: a smile, a turn of the head, steam rising, snow falling, while the picture stays recognisably yours.
Uncensored community LoRAs
Character and style LoRAs built for the 5B model load by link, with no allow-list and no review. Everyone on screen must be an adult.

Prompts

Wan 2.2 5B LoRA prompt examples you can copy

Paste one into the playground as it is, then change one thing at a time.

Text to video
Après-ski in a wingback chair
Result from this prompt
Cinematic video. Blue hour in the lounge of an alpine ski lodge: an adult Nigerian-British woman around thirty with close-cropped natural hair, in an ivory off-the-shoulder knit dress, is curled up in a cognac leather wingback chair, snow drifting past the tall glass behind her. She lowers the book in her lap, tilts her head and gives the camera a slow, smouldering glance. Warm copper lamp light against cold blue dusk, slow push-in, 50mm lens, shallow depth of field.
Why it works. No LoRA. One action and one camera move is the right amount for five seconds.
Try this prompt
Image to video
A look up in the boot room
Result from this prompt
The adult man finishes unlacing the leather ski boot, sits back against the wall, pushes his snow-wet hair back with one hand and looks up at the camera with a slow, easy smile. Steam drifts off his shoulders in the low gold sun. Slow push-in.
Why it works. No LoRA. One action, he sits back and smiles; the steam and the low sun come from the photo.
This mode starts from your own image: add it in the playground before you run.
Try this prompt
Text to video
Wool sweater in the drying room
Result from this prompt
Cinematic vertical video. The drying room of a ski lodge, wool socks and mittens hanging on lines under bare tungsten bulbs, sage-green plank walls. An adult Middle Eastern man in his early thirties with a dark beard and snow in his hair pulls off a knit beanie, runs a hand through his hair, then unbuttons the collar of his navy henley and grins at the camera. Warm light, visible breath, slow handheld push-in.
Why it works. No LoRA, portrait. A handheld push-in and snow on the beanie sell the cold.
Try this prompt

Pricing

Wan 2.2 5B LoRA API pricing

From $0.12 per request.

The playground shows the exact price for your settings before you run. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

See every price tier

Get started

How to use the Wan 2.2 5B LoRA API in three steps

Step 1
Use LoRAs made for TI2V-5B
On Hugging Face, pick LoRAs whose base model is Wan-AI/Wan2.2-TI2V-5B. Files made for the 14B Wan 2.2 models load here without an error and change nothing: in our same-seed test the clip matched the run without a LoRA.
Step 2
Write one moment per clip
Five seconds is short. Name who is in frame, what they wear, one gesture or glance, and the camera move: "she lowers the book and gives the camera a slow glance, slow push-in".
Step 3
Raise the scale step by step
Keep the seed fixed and try 0.5, then 1. In our test a fabric-style LoRA changed the clip a little at 0.5 and more at 1, and three LoRAs stacked changed the face and the room. Put the trigger word first in the prompt.

Open the playground Get an API key

Parameter reference and code samples are in the API and Readme tabs at the top of this page.

Why SpicyAPI

Why run Wan 2.2 5B LoRA on SpicyAPI

No filter from us
We add no content filter of our own. What the model renders is what you get back.
Tested, not guessed
Content labels come from real generations we ran and checked, with the test date on the page.
Pay per result, in dollars
No subscription and no credits. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.
One key for every model
One API key and balance cover image, video, audio and chat models, so you can explore more models from the same account.

FAQ

Wan 2.2 5B LoRA API FAQ

About Wan 2.2 5B LoRA

Wan 2.2 5B LoRA is Alibaba's compact open video model with room for your own weights: up to three LoRA files trained for TI2V-5B per request, each with a strength from 0 to 4, on text-to-video and image-to-video. Every request returns one five-second 720p clip at a flat price. LoRAs are not reviewed and nothing is filtered.