Run models

The best NSFW AI video generators, ranked by test results

By the SpicyAPI team · Updated · Test data as of · How we test

Wan 3.0 is the best NSFW AI video generator we have tested: it delivered all 51 of its test runs across five intensity levels and has the highest Spicy Index of any video model in our data (87.7). The seven models below it trade a little of that range for a better blind-test score, LoRA support or a simpler image-only workflow.

Quick picks

#ModelBest forRuns deliveredBlind testSpicy IndexFromLink
1Wan 3.0Best overall51/5170.987.7$0.045/sec at 480pRun
2Seedance 2.5Best ByteDance model45/4665.882.2$0.074/sec at 480pRun
3MiniMax H3 LoRABest with LoRA styles58/6068.779.6$0.05/sec at 480pRun
4Wan 2.6 FlashBest for unfiltered image-to-video19/1969.278.1$0.022/sec at 720pRun
5Wan 3.0 ProBest blind-test quality43/4372.577.9$0.14/sec at 1080pRun
6Vidu Q3Best Vidu model17/1967.374$0.07/sec at 540pRun
7Seedance 2.0 Fast SpicyBest quick Seedance variant16/1966.572.6$0.081/sec at 480pRun
8Wan 2.5Best older Wan with text input37/3964.167.6$0.045/sec at 480pRun

Runs delivered counts test runs returned as requested across all five intensity levels; softened and refused runs are not counted. “From” is the lowest listed price on SpicyAPI, checked 2026-11-04; higher resolutions cost more. Sample sizes are small for some models — see how we test.

How we picked

The pool was every video model in the SpicyAPI test index with a completed range test: 64 of them on 2026-11-03. We kept one model per product line unless two versions serve different jobs, required at least text or image input, and list the eight survivors in Spicy Index order. Clip tools such as lip-sync and upscalers are left out because they edit footage rather than generate it.

The ranking

Public sample from the Wan 3.0 page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
1

Best overall

Wan 3.0 by Alibaba

  • Highest Spicy Index here
  • Text and image input

No refusals and no softened runs at any level, plus a blind-test score in the top tier of the video board. It takes text, a start image or reference images, and can extend a finished clip, so one model covers a whole project.

Spicy Index
87.7
Runs delivered
51/51
Blind test
70.9
From
$0.045/sec at 480p
  1. Mild 9/9 all delivered
  2. Moderate 9/9 all delivered
  3. Strong 6/6 all delivered
  4. Very strong 9/9 all delivered
  5. Maximum 18/18 all delivered

Delivered every run at all five levels. Inputs: Image to video, Reference to video, Text to video, Video extend.

Try Wan 3.0 on SpicyAPI Full test record

Public sample from the Seedance 2.5 page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
2

Best ByteDance model

Seedance 2.5 by ByteDance

  • Text and image input
  • Small sample: 3 runs or fewer at some levels

Second on the Spicy Index. It softened one run at Very strong and delivered the rest. Pick it when you already know the Seedance look, or when you want reference-to-video with several input stills.

Spicy Index
82.2
Runs delivered
45/46
Blind test
65.8
From
$0.074/sec at 480p
  1. Mild 9/9 all delivered
  2. Moderate 9/9 all delivered
  3. Strong 16/16 all delivered
  4. Very strong 8/9 some held back
  5. Maximum 3/3 all delivered

Clean up to Strong; held back on some runs from Very strong. Inputs: Image to video, Reference to video, Text to video, Video extend.

Try Seedance 2.5 on SpicyAPI Full test record

Public sample from the MiniMax H3 LoRA page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
3

Best with LoRA styles

MiniMax H3 LoRA by MiniMax

  • Loads your LoRA
  • Text and image input

H3 with LoRA loading delivered 58 of 60 runs, with two softened at Very strong. Load up to three style or character weights by URL, the closest hosted equivalent to a local ComfyUI workflow.

Spicy Index
79.6
Runs delivered
58/60
Blind test
68.7
From
$0.05/sec at 480p
  1. Mild 9/9 all delivered
  2. Moderate 9/9 all delivered
  3. Strong 11/11 all delivered
  4. Very strong 11/13 some held back
  5. Maximum 18/18 all delivered

Clean up to Strong; held back on some runs from Very strong. Inputs: Image to video, Reference to video, Text to video, Video extend.

Try MiniMax H3 LoRA on SpicyAPI Full test record

Public sample from the Wan 2.6 Flash page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
4

Best for unfiltered image-to-video

Wan 2.6 Flash by Alibaba

  • No refusals in testing
  • Small sample: 3 runs or fewer at some levels

Every test run came back delivered. It has no text-to-video mode, so start from an image you made or own the rights to.

Spicy Index
78.1
Runs delivered
19/19
Blind test
69.2
From
$0.022/sec at 720p
  1. Mild 3/3 all delivered
  2. Moderate 3/3 all delivered
  3. Strong 3/3 all delivered
  4. Very strong 3/3 all delivered
  5. Maximum 7/7 all delivered

Delivered every run at all five levels. Inputs: Image to video, Reference to video.

Try Wan 2.6 Flash on SpicyAPI Full test record

Public sample from the Wan 3.0 Pro page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
5

Best blind-test quality

Wan 3.0 Pro by Alibaba

  • Best blind-test score here
  • Text and image input
  • Small sample: 3 runs or fewer at some levels

The highest blind-test score on this list, ahead of Wan 3.0, and all 43 test runs delivered. Its Spicy Index still trails Wan 3.0, because the index also weighs public leaderboards and capability.

Spicy Index
77.9
Runs delivered
43/43
Blind test
72.5
From
$0.14/sec at 1080p
  1. Mild 9/9 all delivered
  2. Moderate 9/9 all delivered
  3. Strong 13/13 all delivered
  4. Very strong 9/9 all delivered
  5. Maximum 3/3 all delivered

Delivered every run at all five levels. Inputs: Image to video, Reference to video, Text to video, Video extend.

Try Wan 3.0 Pro on SpicyAPI Full test record

Public sample from the Vidu Q3 page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
6

Best Vidu model

Vidu Q3 by Vidu

  • Small sample: 3 runs or fewer at some levels

Image-to-video only. It softened two of three runs at Very strong and delivered everything else, including all Maximum runs. Its blind-test score is level with Seedance 2.0.

Spicy Index
74
Runs delivered
17/19
Blind test
67.3
From
$0.07/sec at 540p
  1. Mild 3/3 all delivered
  2. Moderate 3/3 all delivered
  3. Strong 3/3 all delivered
  4. Very strong 1/3 some held back
  5. Maximum 7/7 all delivered

Clean up to Strong; held back on some runs from Very strong. Inputs: Image to video.

Try Vidu Q3 on SpicyAPI Full test record

Public sample from the Seedance 2.0 Fast Spicy page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
7

Best quick Seedance variant

Seedance 2.0 Fast Spicy by ByteDance

  • Small sample: 3 runs or fewer at some levels

The Spicy build of Seedance 2.0 Fast scores well above the standard Fast model on range. Start-image input and clip extension only.

Spicy Index
72.6
Runs delivered
16/19
Blind test
66.5
From
$0.081/sec at 480p
  1. Mild 3/3 all delivered
  2. Moderate 3/3 all delivered
  3. Strong 3/3 all delivered
  4. Very strong 0/3 blocked
  5. Maximum 7/7 all delivered

Clean up to Strong; blocked from Very strong. Inputs: Image to video, Video extend.

Try Seedance 2.0 Fast Spicy on SpicyAPI Full test record

Public sample from the Wan 2.5 page on SpicyAPI. It shows the model’s look; it is not one of the test runs counted here.
8

Best older Wan with text input

Wan 2.5 by Alibaba

  • Text and image input

Wan 2.5 is a generation behind 3.0 on quality, but it delivered nearly every run and still accepts a plain text prompt. Good for drafts before you spend on a 3.0 render.

Spicy Index
67.6
Runs delivered
37/39
Blind test
64.1
From
$0.045/sec at 480p
  1. Mild 6/6 all delivered
  2. Moderate 6/6 all delivered
  3. Strong 6/6 all delivered
  4. Very strong 6/6 all delivered
  5. Maximum 13/15 some held back

Clean up to Very strong; held back on some runs from Maximum. Inputs: Image to video, Text to video, Video extend.

Try Wan 2.5 on SpicyAPI Full test record

What separates these models

Two numbers matter for this kind of work. The first is range: how many of our test runs came back as requested, at five intensity levels from Mild to Maximum. The second is quality: a blind-test score from AI judges built by three different companies, who rate each clip from 1 to 10 without seeing which model made it. Each pick shows both alongside the Spicy Index, which combines them with capability and public arena results.

Input type matters too. A text-to-video model can start from nothing; an image-to-video model needs a first frame. Several of the most permissive models on this list are image-only, so check the input column before you choose.

What none of these models will do

Range has a floor. Our tests never involve minors, real or identifiable people, or non-consensual scenarios, and the models on this page are run under terms that forbid all three. A start image of a real person is outside these tests and outside the rules.

Well-known models that held back

Kling and PixVerse are two of the most searched video brands, and H3 HighSpeed is the fast tier of the H3 family, whose LoRA build ranks third above. All of them refused or softened most of the test set, so they are better kept for footage their official apps allow.

  • Kling 4.0 by Kling

    3 of 8 runs delivered. Clean up to Mild; blocked from Moderate. Image to video, Reference to video, Text to video.

    1. 2/2
    2. 0/1
    3. 0/1
    4. 0/1
    5. 1/3
  • PixVerse V6 by PixVerse

    1 of 3 runs delivered. Clean up to Mild; blocked from Moderate. Image to video, Reference to video, Text to video.

    1. 1/1
    2. 0/1
    3. 0/1
    4. —
    5. —
  • MiniMax H3 HighSpeed by MiniMax

    5 of 5 runs delivered. Delivered every run at all five levels. Image to video, Reference to video, Text to video.

    1. —
    2. 1/1
    3. 1/1
    4. 3/3
    5. —
  • Kling 3.0 by Kling

    7 of 15 runs delivered. Clean up to Mild; blocked from Moderate. Character animation, Image to video, Text to video.

    1. 7/7
    2. 0/2
    3. 0/3
    4. 0/2
    5. 0/1

Every scored model is in the filter tracker.

Questions

Which AI video generator allows NSFW content?

In our tests, Wan 3.0, MiniMax H3 with LoRA, Wan 2.6 Flash and Wan 2.5 delivered nearly every NSFW test run. Kling 4.0, PixVerse V6 and MiniMax H3 HighSpeed held back most runs.

Is there a free option?

None of these models is free to run at video resolution; each clip is billed per second or per clip. The model page on SpicyAPI shows the exact price before you start a job.

How often is this ranking updated?

The scores come from the SpicyAPI test index. We rebuild this page when the index changes, and the date at the top shows the data we used.

Open the model picker →