Two NSFW-capable ByteDance video models, with the content filters relaxed and the prompt rewriting removed. One reaches 30 seconds with synchronized audio, the other renders 1080p with a seed you can pin. Both sit behind the same Wiro API as everything else in the catalog.
These are not two versions of the same thing. One buys you length and audio, the other buys you resolution and repeatability. Pick by whichever constraint binds your shot.
The newest of the two. Generates 4 to 30 seconds at any whole second, with audio produced alongside the visuals rather than dubbed on afterwards.
Shorter clips, sharper frames. The only one of the two that renders 1080p, exposes a seed, and can lock the camera in place.
Genuine sample output taken from each model page. Hover to play. Captions describe what the model can do, not the prompt behind each clip.
Clips run to 30 seconds in a single request, at any whole second from 4 upward.
Up to 30s · Native audioThe only one of the two that renders 1080p, with lip-sync on generated speech.
1080p · Lip-syncAudio is generated in the same pass as the visuals, so it lands on the action.
Synced audio · 720pSet a first frame and an optional last frame, and the model fills the middle.
First and last frameExport as mov to hold higher colour precision through grading and compositing.
mp4 or mov outputPin the seed and a rerun reproduces the same take.
Seed control · 1080pLock the camera in place for a static frame that will not drift mid-take.
Fixed cameraFive or ten second clips, from text or a first frame.
5 or 10s · Native audioNeither wins outright. Length and audio sit on one side, resolution and repeatability on the other.
| Capability | Seedance 2.5 Uncensored | Seedance Pro v1.5 Uncensored |
|---|---|---|
| Clip length | 4 to 30 seconds, any whole second Longer | 5 or 10 seconds only |
| Resolution | 480p, 720p | 480p, 720p, 1080p Sharper |
| Synchronized audio | Yes, on or off | Yes, with lip-sync |
| Seed control | Not exposed | Yes, for repeatable runs Wins |
| Fixed camera | Not exposed | Yes Wins |
| First / last frame | Yes | Yes |
| Aspect ratios | Adaptive plus six fixed, to 21:9 | Adaptive plus six fixed, to 21:9 |
| Output format | mp4 or mov New | mp4 |
| Watermark | Off by default | Off by default |
| Price per request | $0.42 to $6.94 | $0.06 to $1.18 Cheaper |
Read from the live Wiro model catalog. Only successful runs are billed, so a rejected or failed generation costs nothing.
Six situations, and the model that handles each without a workaround.
Pro v1.5 gives you 5 or 10 seconds and nothing between. Anything longer has one answer.
Seedance 2.5 tops out at 720p. If the frame has to hold up on a large screen, take the shorter clip.
Pin the seed and a rerun reproduces the take. Seedance 2.5 does not expose one, so identical inputs still vary.
Both generate audio in the same pass, but 30 seconds of synced sound is a different product from 10.
Fixed camera plus lip-sync covers pieces to camera without the frame drifting mid-take.
Request mov output to hold higher colour precision through keying, compositing and colour work.
The standard Seedance releases run a prompt enhancement pass that rewrites your text into a brandless, filter-safe version before generation. It helps most users and quietly rewrites the shot for the rest. The uncensored variants drop that step, so what you typed is what gets rendered. Capability, speed and the request shape are otherwise unchanged, and your own product rules, ByteDance terms and local law still apply, so build the guardrails your use case needs at your layer.
Post your parameters and Wiro returns a task token immediately. Poll /Task/Detail or pass a callbackUrl and let the webhook find you. Swapping between the two models is a slug change.
import requests headers = { "x-api-key": API_KEY, "Content-Type": "application/json", } run = requests.post( "https://api.wiro.ai/v1/Run/bytedance/" "seedance-2-5-uncensored", headers=headers, json={ "prompt": "your prompt, unrewritten", "duration": "30", "resolution": "720p", "ratio": "16:9", "generateAudio": "true", "outputFormat": "mov", "watermark": "false", } ).json() task = requests.post( "https://api.wiro.ai/v1/Task/Detail", headers=headers, json={"tasktoken": run["socketaccesstoken"]}, ).json() # pexit == "0" means success print(task["outputs"])
Most of this is ordinary production work that a filter-safe rewrite quietly alters: a brand name dropped, a weapon removed from a period scene, a medical detail softened.
Every model page has a playground, so you can see which one renders your shot before writing a line of code. Same request shape across the catalog, and only successful runs are billed.