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 2.5 Review: The Cheap Way to Get Sound With Your Clip

Pick Wan 2.5 for 5-to-10-second clips with generated sound from text or a still; step up to Wan 2.6 only for multi-shot cuts or reference casting.

Create in StudioOpen the Playground

Spicy Index

67.6/ 100

Freedom Score

99.0/ 100

Content levels

Tested 6 of 6 probe types · 39 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
$0.45 per 5-second clip
Median time
1.6 min
Success rate
Collecting data

Wan 2.5 was Alibaba's first Wan release to ship native audio, and it's still less expensive than its own multi-shot sibling Wan 2.6 for a clip with sound baked into the same file. From $0.045 a second at 480p, either endpoint returns 5 or 10 seconds at up to 1080p, with the soundtrack timed to the picture rather than added afterward. In our latest arena batch it rendered every rung of our content ladder we ran on both text and image input, and only softened two of the toughest L5a text prompts (consensual adult restraint and impact play).

What Wan 2.5 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

    Early morning at a stone harbour wall. An adult woman in a navy one-piece swimsuit and a white swim cap stands at the edge, shakes out her arms, then dives cleanly into flat green water and surfaces laughing, pushing her goggles up. Gulls lift off the wall behind her. Low handheld camera at water level, soft pink dawn light, mist on the water. Audio: gulls, a splash, her breath and a short laugh, distant boat engine.

    Try this model

  • Image to video

    She shifts her weight, turns her shoulders toward the lens and takes two slow steps forward, the satin of the dress catching the light as it moves. She tucks her hair behind one ear and holds the look. Locked camera, clean grey studio, soft footsteps on the floor.

    Try this model

  • Text to video

    Ten seconds, jazz bar: a saxophone solo under a single overhead lamp, tie pulled loose, sweat on the throat, the horn's sound matching the fingering exactly. Cinematic, shallow depth of field, heavy film grain, smoke blue and brass.

    Try this model

Make your own with Wan 2.5

Create in StudioOpen the Playground

The verdict

Pick it when the job is a short, sound-bearing clip and you don't need multiple camera angles inside one call. Step up to Wan 2.6 only when the brief genuinely needs multi-shot cuts or reference-video casting; otherwise 2.5 does the same job for half the price.

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

What we liked and what held it back

What we liked

  • Sound comes free, and it's timed. Both tasks render generated audio alongside the picture at no extra charge, and audio_url can lock the motion to a recorded line, a piece of music or a sound effect. In our blind arena runs, judges gave its lip sync on a close-up delivery line noticeably higher marks than on several of the other clips we compared it against.
  • Priced well below Wan 2.6 for sound. $0.045 a second at 480p, less than half of base Wan 2.6's $0.09, with 720p and 1080p available on the same schema. A 5-second clip at 480p is $0.225.
  • It holds a painted texture. Fed a painting as the opening frame and asked to bring the boats and clouds to life, it kept the brushstroke texture rather than flattening it into a photo, while the water and sky picked up slow, believable motion.
  • A fixed seed for consistent product turns. The macro bottle-turn example locks the label and silhouette while only the camera and the light move — useful for repeatable product shots.
  • Full content ladder on both endpoints. Every rung we tested renders from a still and from plain text, so you don't lose freedom by switching input type.

What held it back

  • Hands and hand-held props still wobble. On a guitarist prompt, our judges flagged "heavily deformed hands that morph unnaturally" mid-strum; on a wok-toss prompt the chef was mostly cropped out of frame. If a shot leans on close hand work, plan a couple of extra takes.
  • Physics-heavy single actions get abbreviated. Asked for a wine glass sliding off a table edge and shattering on stone, it showed mostly the aftermath rather than the fall and the impact — the event that matters most in the prompt is the one it's most likely to skip.
  • Two of our toughest text prompts came back softened. Everything else in our ladder rendered; two L5a prompts submitted as plain text were toned down rather than shown as written. The same scenario from an image opening frame rendered in full, so a still is the more reliable path for the hardest scenes.
  • 10 seconds is the ceiling, in one take. For anything longer, plan to chain clips from the last frame rather than asking for one continuous shot.
  • No multi-shot or reference casting. For a numbered shot list cut inside a single call, or for carrying a face or an object out of existing footage, that's Wan 2.6.

What the clips look like

The gallery holds published, safe-for-work examples from the Wan 2.5 model page.

A dawn dive. A swimmer at a stone harbour wall shakes out her arms, dives cleanly and surfaces laughing; the splash and the following ripple sound track lands roughly where the water breaks.

A satin turn. From a still portrait, ten seconds at 1080p: she turns her shoulders, takes two steps, and the negative prompt keeps jewellery and extra people from drifting into frame.

A ten-second sax solo. A single continuous take in a jazz bar; judges praised the close-up facial detail here even while marking the fretting-style hand work on a separate guitar prompt as its weak point.

Off camera, in our latest same-prompt arena batch, the sound-first prompts we ran (a cafe conversation, a tearful confession, a wine glass shattering, a wok fire) showed the same split every time: dialogue and emotion land well, and the harder the prompt asks it to lean on precise hand or object physics, the more it starts to guess.

Prompts that worked

Name the sound source in the same sentence as the action, and Wan 2.5 will usually time the two together without extra instruction.

1. Sync sound to a visible source.

text
A wine glass slides off the edge of a wooden table and shatters on a stone floor, with a sharp crash and the tinkle of scattering shards, in a quiet kitchen.

Run this prompt on Wan 2.5 →

2. Keep a portrait's face across a turn (image-to-video).

text
She shifts her weight, turns her shoulders toward the lens and takes two slow steps forward, the satin of the dress catching the light as it moves. She tucks her hair behind one ear and holds the look. Locked camera, clean grey studio, soft light.

Run this prompt on Wan 2.5 image-to-video →

Weak: "A woman dances in a room." No sound cue, no named action for the audio track to follow. Strong: name the object or voice making the sound, and where it happens in the shot, as in prompt 1.

How far it goes

Wan 2.5 rendered every rung we ran on both text and image input, from L1 Suggestive scenes through L3 Strong content, L4 Intense content and L5 Extreme prompts (L5a consensual adult restraint and impact play, L5b cinematic gore). The one exception: two of five L5a prompts submitted as plain text came back softened rather than shown as written; the same scenario from a still image rendered in full each time we tried it. The page label is Uncensored: we add no content filter on our side. Full pass rates, by rung and by input type, are on the test report for this model.

Adults only, in every prompt, still and output. A real, identifiable person needs to have consented, and you need the rights to any photo you use as an opening frame.

What the community says

Wan 2.5 was the release that took Alibaba's Wan line out of open weights, and the community noticed the same day. On GitHub, an issue was opened against the still-open-source Wan 2.2 repository titled "Thank you Alibaba for deceiving and using the open source community", arguing that the earlier open releases had mostly served as marketing for the closed models that followed — 2.5 among them. The issue remains open.

Our own runs don't speak to the open-weights question, but they do confirm the trade the community was reacting to: 2.5 is API-only, and it's also the version that added native audio, which the earlier open Wan 2.2 never had.

Alternatives

  • The same idea with multi-shot cuts? Wan 2.6 adds numbered shot lists and reference-video casting in one call, at double the per-second price.
  • Even cheaper, no sound? Wan 2.6 Flash drops audio and text-to-video but costs about a quarter of Wan 2.6's rate. Our Wan 2.6 Flash review.
  • Open weights you can run yourself? Wan 2.2 is Alibaba's last openly released version; it has no native audio and holds back the toughest content rungs compared with 2.5.
  • A different family entirely? MiniMax H3 also renders sound and every content rung we tried, at a different price curve. Our MiniMax H3 review.

How far Wan 2.5 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 videoText to video
MildThe lightest prompts in the test
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
StrongMid-scale; must be tested before a model is ranked
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
Very strongWhere most models start to hold back
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 B
Class A: 4/4 deliveredClass B: 3/3 delivered20.0 / 20 pts
Class A: 3/5 deliveredClass B: 3/3 delivered18.0 / 20 pts
Task score100.0 / 100.0Tested Sep 25, 202698.0 / 98.0Lowest: pulls the average downTested Sep 25, 2026

Tested Sep 16, 2026 to Sep 25, 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 2.5 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
51.1
Blind output test 35% of quality
64.1
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
58.5
Freedom Score 30% weight
99.0
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 clip10 s
6.0 / 20
Native audioYes
15.0 / 15
Reference inputsNo reference images · no video or audio
0.0 / 15
Frame controlFirst frame only
5.0 / 10
Task coverageImage to video, Text to video, Video extension
7.5 / 10
Customization (LoRA)No LoRA support
0.0 / 5

Arena rating: 944 (95% interval 836–1054); votes are still coming in.

What Wan 2.5 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 videoText to videoVideo extension
Resolutions480p · 720p · 1080p480p · 720p · 1080p480p · 720p · 1080p
Clip length5–10 s5–10 s5–10 s
Aspect ratios—16:9 · 9:16 · 1:1 · 4:3 · 3:4—
Native audioYesYesYes
Seed controlYesYesYes
Negative promptYesYesYes
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
1.6 min
90th percentile
2.3 min
Wait per second of video
10.8 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.

What people say about Wan 2.5

0 positive · 0 mixed · 1 negative

  • Thank you Alibaba for deceiving and using the open source community

    GitHub· Sep 24, 2025· NegativeSource

Quotes are verbatim from public posts, linked to the original. They are not ratings and don’t feed any score.

Models to consider instead of Wan 2.5

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 2.5 FAQ

Try Wan 2.5 yourself

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

$0.45 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 dark room with a wall of tall windows over a glittering city at night, a single small figure at the glass

    Wan 3.0 Prime

    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.

    Spicy Index
    75.8
    Freedom Score
    80.7

    Tested Sep 24, 2026

    Read the review