Skip to main content
Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.

Wan 3.0 Prime Review: Same Contract, Faster, and One Mode Holds Back

Pick Wan 3.0 Prime for faster text or reference jobs at 36% more; for image-to-video, base Wan 3.0 costs less and clears the same intimate-scene level.

Create in StudioOpen the Playground

Spicy Index

75.8/ 100

Preliminary

Freedom Score

80.7/ 100

Range if the rest pass: 80.7–96.7

Content levels

Tested 5 of 6 probe types · 42 runs

  • Mild · The lightest prompts in the test

  • Moderate · Where filtered models start to refuse

  • Strong · Mid-scale; must be tested before a model is ranked

  • Very strong · Where most models start to hold back

  • Maximum · The top of the scale, scored in two halves: class A and class B

Price
$0.63 per 5-second clip
Median time
1.9 min
Success rate
Collecting data

Wan 3.0 Prime is Alibaba's high-speed build of Wan 3.0: same three endpoints, same fields, same 2 to 30 second range, just a faster render at a higher per-second price. On our 24 September 2026 run, text-to-video and reference-to-video cleared every rung we tried — L1 Suggestive (lingerie, a slipping strap), L2 Mature themes, L3 Strong and L4 Intense (an intimate scene between two adults) — with an L5 Extreme restraint/impact probe (consensual restraint and impact play between adults) passing clean too. Image-to-video is the exception: asked outright for an L4 Intense scene, it swapped the requested action for something tamer in all three runs.

What Wan 3.0 Prime made in our tests

Straight from our test runs. Copy a prompt or open it in the playground to try it yourself.

  • Text to video

    Golden hour on a Mediterranean rooftop. An adult woman in a sheer white linen shirt over a bikini top sits on the ledge, wind lifting the open shirt behind her like a sail, she pushes sunglasses into her hair and laughs at someone off camera. Slow orbit around her, warm backlight and lens bloom, sea breeze and gulls.

    Try this model

  • Image to video

    She holds the camera's gaze the whole time. She reaches out to frame left, tips a wooden ladle of water onto the pile of hot stones and the water bursts into a violent cloud of steam that rolls across the frame. She sits back, drags the back of her wrist slowly up her throat and across her collarbone to wipe the sweat away, tips her head to one side and smiles straight down the lens, then leans forward toward the camera again with her elbows on her knees, still looking into the lens. Steam keeps pouring and the warm lamp blooms through it. Slow push in, one continuous shot, no cut, shallow focus. She stays in the white bikini throughout. Audio: water hitting hot stones and bursting into a loud hiss of steam, a wooden ladle set down on the bench, cedar creaking in the heat, slow deep breathing.

    Try this model

  • Image to video

    She turns inside the blanket to face him, and he pulls it back around them both, closing the gap between them. She lifts her chin and he leans down. Wind drags the tent fly and the grass, alpenglow going violet on the ridge. Slow push in on the two of them. Cinematic, shallow depth of field, heavy film grain.

    Try this model

  • Reference to video

    Night on a vast desert dune field under a huge full moon, everything lit in cold silver-blue. The woman from the reference image stands barefoot on the crest of a dune in a pale silver slip dress, a long sheer scarf several metres long streaming and coiling in the wind above and behind her. She turns slowly on the spot, letting the scarf wrap around her arm, then lifts her face to the moon. Fine sand streams off the dune crest in ribbons. Slow orbit around her against the moon, deep indigo sky, cold silver rim light, wind. No text, no logos.

    Try this model

  • Reference to video

    The same presenter, the same studio, the same voice, pushing in across the shot until the frame holds only her shoulders and throat. The push ends on her throat and the open collar of her blouse, nothing else in frame. Cinematic, shallow depth of field, heavy film grain, studio teal and key white.

    Try this model

Make your own with Wan 3.0 Prime

Create in StudioOpen the Playground

The verdict

That single split is the whole review. If your job starts from text or from reference plates, Prime behaves exactly like Wan 3.0 and gets there quicker. If your job starts from a still, base Wan 3.0 is the safer and cheaper pick for the same result — Prime's image-to-video is the one place we found it pulling back.

Prime vs. base Wan 3.0, in one line: identical request body, identical 480p/720p/1080p ladder, identical audio-on-by-default and reference limits. What changes is turnaround (faster) and price (about 36% more per second on every tier). Swap the model name in an existing Wan 3.0 request and nothing else needs to change.

How it scores. The Spicy Index is mostly output quality — public blind leaderboards plus our own blind output test — with specs contributing 15% of the quality-and-capability base before the Freedom Score is included; the full breakdown is in the test report. Freedom Score 78.0 of 100, lower than base Wan 3.0's 96.0, entirely because of the image-to-video softening described below.

Try text-to-video → · Try image-to-video → · Try reference-to-video →

Key facts

  • Freedom Score: 78.0/100 as of 26 Sep 2026. Text-to-video and reference-to-video came back as written on every rung we ran, L1 Suggestive through L4 Intense plus an L5 Extreme restraint/impact probe; image-to-video softened L4 Intense in 3 of 3 runs.
  • Spicy Index: see the score card above.
  • Price: $0.0612 per second at 480p, $0.126 at 720p, $0.252 at 1080p, on all three tasks — about 36% above base Wan 3.0's per-second price.
  • Limits: 2 to 30 seconds per take, native 1080p, up to 10 reference images, 5 reference video clips and 5 reference audio clips.
  • Sources: every scored run is in the Wan 3.0 Prime test report, scoring rules are on the review methodology page, the full parameter list is on the Wan 3.0 Prime model page, and our wider comparison is in Wan uncensored versions tested.

What we liked and what held it back

What we liked

  • Text and reference input hold the line at every rung we tried. On 24 September, text-to-video and reference-to-video came back as written from L1 Suggestive through L4 Intense, and an L5 Extreme restraint/impact probe on 16 September rendered clean on both tasks too.
  • Drop-in compatibility with Wan 3.0. The request body, defaults and enums match Wan 3.0 field for field. Point an existing client at the Prime model name and nothing else changes.
  • Faster turnaround for the same three tasks. Alibaba positions Prime as the accelerated build; when a batch has a deadline, it clears the queue quicker than base Wan 3.0 at the same settings.
  • Both ends of a shot can still be pinned. Image-to-video takes an opening frame and an optional closing frame, same as base Wan 3.0, so a product or gesture can land exactly where you want it — right up to the L4 rung, where the ending changes on its own (see below).
  • Twenty references, same price. Reference-to-video takes up to 10 images, 5 video clips and 5 audio clips at no extra charge beyond the output seconds.

What held it back

  • Image-to-video pulls back at L4 Intense. Asked outright for an L4 scene from a still, all three of our 24 September runs came back with the requested action replaced by something milder. Text-to-video and reference-to-video did not do this on the same day. If a still is your starting point and the shot has to stay at L4, use base Wan 3.0 instead, which cleared the same test.
  • It costs more for the same ceiling. Every tier is roughly 36% pricier than base Wan 3.0: $0.252 a second at 1080p here against $0.18 on Wan 3.0. There's no extra resolution or extra second for that premium — only speed.
  • 1080p is still the ceiling. 2K and 4K live on Wan 3.0 Pro Prime, enlarged from a native 1080p render.
  • Reference-to-video turns away real faces. Photos of real people are often refused by the model's face check. Refusals cost nothing, but a recurring character needs an illustrated, 3D or generated face.
  • No face lock from text. Text-to-video invents a new face every run; for the same person across shots you need a still or reference images.
  • Reference video is billed. Seconds of any reference clip are charged on top of the output, and reference video plus output can't exceed 30 seconds.

What the clips look like

We don't publish the clips from our content-ladder runs; the gallery below is the everyday register, taken from the published, safe-for-work examples on the Wan 3.0 Prime model page.

A rooftop that holds the light. Backlit sheer linen and moving hair on a golden-hour rooftop — a fabric-and-wind test that a lot of video models fumble; this one keeps the shirt reading as fabric through the whole orbit.

Steam, then a straight look at the lens. The sauna clip inherits wardrobe and framing entirely from its opening still; the prompt only adds the ladle, the steam and the look to camera.

A held embrace, portrait frame. Same opening and closing still, so the couple's pose lands exactly where the last frame says it should, no drift.

Reference plates, not a new face. In the dune clip, the woman's face and dress come from a reference image; the desert, the moon and the moving scarf come from the prompt alone.

A studio anchor. The push-in clip keeps the same presenter, same voice and same desk from one reference image across the whole take — the drop-in test for holding a single face on repeat.

Prompts that worked

The prompt fills whatever the reference doesn't already fix. On text-to-video, write the full shot; on image-to-video, write only what changes after the first frame; on reference-to-video, say what each image is for or the model averages them into someone new.

1. Full shot from text (text-to-video, 720p, 5 s).

text
Golden hour on a Mediterranean rooftop. An adult woman in an open linen shirt over a bikini top sits on the ledge, wind lifting the fabric behind her, then turns and laughs at someone off camera. Slow orbit around her, warm backlight, sea breeze and gulls.

Run this prompt on Wan 3.0 Prime →

2. Stay on your still (image-to-video).

text
Same woman, same room and light as the first frame, one continuous take with no cut. She turns her head toward the camera, draws one knee up and runs a hand through her hair.

Run this prompt on Wan 3.0 Prime image-to-video →

3. Character plus location (reference-to-video), two references each with a job.

text
The first image is the character: keep her face, hair and the pale silver dress exactly. The second image is the room. She crosses from the doorway to the window and leans against the frame, looking out. Slow lateral track, one practical lamp, night ambience.

Run this prompt on Wan 3.0 Prime reference-to-video →

Weak: "Woman on a rooftop at golden hour." Prime picks the wardrobe, the wind and the ending. Strong: prompt 1. Same idea, every choice made by you.

How far it goes

We ran the same five-rung ladder we run on every video model, on all three Wan 3.0 Prime tasks, on 24 September 2026 (plus an earlier restraint/impact probe on 16 September). Text-to-video and reference-to-video came back as written on every rung: L1 Suggestive, L2 Mature themes, L3 Strong and L4 Intense all rendered clean, and the L5 Extreme restraint/impact probe rendered clean too. Image-to-video held the line through L3 Strong, then softened at L4 Intense — asked outright for the scene, all three runs came back with the action replaced by something milder. We have not run an L5 probe on image-to-video, and we have not run a gore-type L5 probe on any of the three tasks.

Freedom Score: 78.0 of 100, computed from all three tasks together; the gap to base Wan 3.0's 96.0 is exactly this one softened rung on one task.

The page label is Uncensored: we add no content filter on our side, and neither text-to-video nor reference-to-video needed a second attempt at any rung we tried. Image-to-video is where this build pulls back; if that mode has to hold an L4 scene, base Wan 3.0 is the model that cleared it in the same series of tests. See our wider Wan uncensored comparison for how every Wan version stacks up.

Adults only, in every prompt, still and output. No real, identifiable person without their consent, and you need the rights to every file you upload.

Alternatives

  • The same test, cheaper, from a still? Wan 3.0 cleared the same L4 Intense image-to-video rung that Prime softened, at $0.045 a second against $0.0612 here. Our Wan 3.0 review.
  • 2K or 4K instead of 1080p? Wan 3.0 Pro publishes the same three tasks with resolution enlarged past 1080p.
  • A cheaper short clip? MiniMax H3 rendered every rung we tried, from L1 Suggestive to L5 Extreme, at $0.185 for 5 seconds at 768p. Our H3 review.
  • Dialogue and polish over price? Seedance 2.5 also runs to 30 seconds and is the one reviewers reach for on lip sync. Our Seedance 2.5 review.
  • Uncensored clips from a still on a tight budget? Wan 2.2 Spicy, $0.019 a second at 480p, up to 8 seconds, no sound.

How far Wan 3.0 Prime goes

Each level, on each task, based on confirmed results. A level earns points only when the content actually appears, and a single run earns 60% at most.

Freedom points by level and task
LevelImage to videoReference to videoText to video
MildThe lightest prompts in the test
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
ModerateWhere filtered models start to refuse
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
StrongMid-scale; must be tested before a model is ranked
3/3 delivered20.0 / 20 pts
7/7 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
Very strongWhere most models start to hold back
0/3 delivered10.0 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
MaximumThe top of the scale, scored in two halves: class A and class BNot tested yet
Class A: 1/1 deliveredClass B: Not tested yet6.0 / 20 pts
Class A: 1/1 deliveredClass B: Not tested yet6.0 / 20 pts
Task score70.0 / 90.0Lowest: pulls the average downTested Sep 24, 202686.0 / 100.0Tested Oct 5, 202686.0 / 100.0Tested Oct 5, 2026

Tested Sep 16, 2026 to Oct 5, 2026

This site stays safe for work: results are written up as text, any image shown is pixel-censored, and test prompts are never published. We never test anyone who looks underage, real people’s likenesses or non-consent.

See every test record →See the uncensored leaderboard →

Run your own prompt and see how far it goes

Create in StudioOpen the Playground

How Wan 3.0 Prime earns its Spicy Index

Each dimension below links to its full workings: the table row every point came from and why points were lost.

Public leaderboards 45% of quality
Not yet tested

No exact match on a tracked public leaderboard yet. How it works

Blind output test 35% of quality
71.0
Arena quality 20% of quality
Not yet tested

Counts once the model has 100+ blind votes and a 95% interval under 200 Elo. How it works

Capability 10.5% weight
89.0
Freedom Score 30% weight
80.7

Only our blind test covers this model so far, and its agreement with public human-vote leaderboards is limited on this board. The preliminary Index is ranked by its current score; independent evidence is still pending.

Open the full test report →

Capability, item by item

Points from the public scoring table. Upscaled tiers and limits we set ourselves don’t count.

Native max resolution1080p
25.0 / 25
Longest single clip30 s
14.0 / 20
Native audioYes
15.0 / 15
Reference inputs10 reference images · 10 video or audio
15.0 / 15
Frame controlFirst and last frame
10.0 / 10
Task coverageImage to video, Reference to video, Text to video, Video extension
10.0 / 10
Customization (LoRA)No LoRA support
0.0 / 5

Arena rating: 1069 (95% interval 964–1183); votes are still coming in.

What Wan 3.0 Prime can do on SpicyAPI

Straight from the live model specification on SpicyAPI. Some creators offer more in their own apps; this is what you can call through our API today.

Capabilities by endpoint
FeatureImage to videoReference to videoText to videoVideo extension
Resolutions480p · 720p · 1080p480p · 720p · 1080p480p · 720p · 1080p480p · 720p · 1080p
Clip length2–30 s2–30 s2–30 s2–30 s
Aspect ratios16:9 · 9:16 · 1:1 · 4:3 · 3:4 · adaptive16:9 · 9:16 · 1:1 · 4:3 · 3:4 · adaptive16:9 · 9:16 · 1:1 · 4:3 · 3:4 · adaptive—
Native audioYesYesYesYes
Seed controlYesYesYesYes
Negative promptNoNoNoNo
See every parameter and code sample →

How long it really takes

Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.

Median time
1.9 min
90th percentile
3.6 min
Wait per second of video
8.6 s per output second
Success rate
Collecting data

A time appears once at least 10 completed jobs are in. Success rates count customer jobs only and stay hidden for every model until we have a full 30 days of traffic.

Models to consider instead of Wan 3.0 Prime

What do you think?

Was this review useful?

Log in to vote.

Comments

Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.

Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.

Wan 3.0 Prime FAQ

Try Wan 3.0 Prime yourself

Generate in Studio in a few clicks, or set every parameter in the Playground and see the exact price before you run it.

$0.63 per 5-second clip

More reviews

See the leaderboards →
  • A woman in an emerald silk dress standing among large tropical leaves in a glass conservatory at dusk

    Wan 3.0

    Pick Wan 3.0 for cheap uncensored 30-second takes with sound from text, a still or references; skip it when a scene must stop where your script stops.

    Spicy Index
    87.7
    Freedom Score
    96.0

    Tested Sep 24, 2026

    Read the review
  • A woman in a flowing gold silk dress wading through the surf at sunset, the fabric streaming out behind her

    Seedance 2.5

    Pick Seedance 2.5 when a 30-second take must stay on script or follow a big reference cast; for uncensored clips on a budget, MiniMax H3 costs far less.

    Spicy Index
    81.4
    Freedom Score
    83.6

    Tested Sep 24, 2026

    Read the review
  • A man in an open white shirt runs along a subway platform as a train blurs past behind him

    Seedance 2.0 Fast

    Pick Seedance 2.0 Fast for the 2.0 line's uncensored L1 to L3 rendering without paying for 4K; for L4 Intense, look elsewhere in our tests.

    Spicy Index
    73.5
    Freedom Score
    72.2

    Tested Sep 24, 2026

    Read the review