Wan 3.0 vs Seedance 2.5

Both can hold a 30-second shot. Seedance 2.5 takes a much larger pile of reference images, video, and audio. Wan 3.0 on this site is a credit-priced browser generator with text, a first frame, and an optional last frame. Numbers for Seedance are from ByteDance and Volcengine, as of Oct 2026.

30s on both Seedance takes video references Different price units

Try Wan 3.0

Where each one is the simpler choice

Closer match on length

A single generation tops out at 30 seconds on both, as of Oct 2026. Seedance starts at 4 seconds and can extend. Wan 3.0 here starts at 2 seconds and stops at the length you pick.

Seedance takes more references

Up to 30 images, 10 videos, and 10 audio clips in one Seedance pass, plus timestamp editing in the official write-up. This Wan form is text, a first frame, and an optional last frame.

Different price units

Seedance on Volcengine is token priced, with published yuan-per-second examples. Wan 3.0 here is credits that scale with length and resolution, not a single dollar-per-second rate.

Wan 3.0 and Seedance 2.5, side by side

Competitor numbers come from ByteDance Seed and Volcengine Ark docs fetched for this page. Where a per-second figure depends on the input mix, the cell says so.

Feature	Wan 3.0 on wan-ai.app	Seedance 2.5
Max duration	2–30 seconds. Any whole second in that range. This generator does not offer automatic duration.	4–30 seconds, or -1 so the model picks a length inside the valid range. Single pass up to 30 seconds, with multi-round extension described by ByteDance. As of Oct 2026.
Resolution	480P, 720P, and 1080P.	480p (8-bit), 720p (8-bit), and 1080p (10-bit, H.265). 4K is not in the Seedance 2.5 output list on the Volcengine model page. As of Oct 2026.
Native audio	On or off for each clip. Dialogue, ambience, and effects when it is on.	Supported. The Volcengine model page lists generated audio as available and shows a generate_audio flag. ByteDance describes 30-second audio-video clips. As of Oct 2026.
Reference inputs	Text. First frame, plus an optional last frame. Templates can send 1–3 reference photos. Reference video and reference audio are not accepted on this site.	Up to 30 images, 10 video clips, and 10 audio clips in one pass. Reference video and reference audio are each capped at 30 seconds total. As of Oct 2026.
Price per second	Not a flat public dollar-per-second rate. Standard credits: 1 per 4 seconds at 480P, per 2 seconds at 720P, and per 1 second at 1080P, rounded up, minimum 1. Prime is a faster model on the same form and costs more. Pack prices are on the pricing page.	Not one flat rate for every job. Volcengine's worked example for 16:9, 5 seconds, no video input: ¥0.67/s at 480p, ¥1.51/s at 720p, ¥3.74/s at 1080p. Token price is ¥70 per million at 480p/720p without video input and ¥42 with video; ¥77 and ¥46 at 1080p. Page updated 2026-10-24.

Figures below the Wan column describe this site as of Oct 2026. Competitor cells are quoted from the sources list and marked when a number is not publicly listed.

Use Wan 3.0 here when

You want a browser generator with a known credit cost, clips anywhere from 2 to 30 seconds, and 1080P without bringing your own reference video. Product shots, text prompts, and a first frame are the everyday path.

Use Seedance 2.5 when

The job needs a stack of reference images, a reference video, or a reference audio clip, or you want the editing and extension flow ByteDance describes. Budget the Volcengine token price, which is published in yuan and changes with resolution and whether video is in the input.

Frequently asked questions

Does Seedance 2.5 generate longer clips than Wan 3.0?

Both can do a 30-second clip in one pass, as of Oct 2026. Seedance 2.5's published range is 4–30 seconds, or -1 to let the model pick a length. Wan 3.0 on this site is any whole second from 2 to 30, and this generator does not offer an automatic length. Seedance also documents multi-round extension. Wan 3.0 on this site does not.

Which one accepts more reference files?

Seedance 2.5. ByteDance and the Volcengine model doc say one pass can take up to 30 images, 10 video clips, and 10 audio clips. On wan-ai.app the main Wan 3.0 form takes a text prompt, a first frame, and an optional last frame. Reference-photo templates send one to three stills. Reference video and reference audio are not accepted here.

Is Seedance 2.5 priced per second?

Volcengine bills doubao-seedance-2.5 in tokens, then publishes example yuan-per-second figures for a 16:9, 5-second clip with no video input: ¥0.67 at 480p, ¥1.51 at 720p, and ¥3.74 at 1080p, updated 2026-10-24. A clip that includes input video uses a different token rate, so there is not one flat per-second price for every job.

Is wan-ai.app the Seedance or Wan product site?

No. wan-ai.app is an independent generator. It is not ByteDance, not Volcengine, and not Alibaba's create.wan.video.