A week apart, two Chinese video models both hit 30 seconds
ByteDance's Seedance 2.5 and Alibaba's Wan3.0 shipped within days of each other, both generating 30-second clips in one pass. On the surface it looks like a head-on collision. The strategies couldn't be more different.
One leans hard into professional film. The other digs into office workflows. If you make e-commerce shorts or teaching clips, picking wrong costs you thousands a year.
The cards, laid out
- Seedance 2.5 (ByteDance, released July 31; live on Jimeng/Doubao, Volcano Engine API opened August 7) 30-second native output, up to 4K, up to 50 reference assets per prompt (30 images + 10 videos + 10 audio), second-level timestamp editing and region-level inpainting. Priced per token: 42 yuan per million tokens with video input, 70 yuan per million for text-to-video, about 50% higher than the previous generation. At 720p a second runs roughly 1.5 to 2.1 yuan, so a 30-second clip lands around 45 to 63 yuan.
- Wan3.0 (Alibaba, public beta August 7; on Alibaba Cloud Bailian, Wanxiang, Qianwen) 30-second native output, up to 1080p, and the only one that turns doc/xls/ppt/pdf/md files directly into video (50 pages max). Priced per second: 0.3 yuan at 480p, 0.6 at 720p, 1.2 at 1080p, so a 30-second 1080p clip is about 36 yuan.
What the numbers don't tell you
Seedance 2.5 wins on ceiling. 4K, shot continuity, region editing, green-screen and white-model references make it fit for ad previews and film-grade output. Wan3.0 caps at 1080p, but "drop a PPT in, get a video out" is a real strength - courseware, product manuals, annual reports go in with a prompt and a demo video comes out.
The cost gap is clear. Same 30 seconds: Wan3.0 at 1080p is about 36 yuan, Seedance 2.5 about 45 to 63, and per-second billing is more predictable than per-token for a small team.
Each sits inside its own ecosystem: Seedance with ByteDance (Jimeng/Doubao/Volcano), Wan3.0 with Alibaba (Bailian/Wanxiang/Qianwen).
Which one to open
- Pick Wan3.0 if you ship e-commerce volume, courseware, or internal reports on a tight budget and want office documents as input.
- Pick Seedance 2.5 for ad previews, film-grade clips, 4K, or region-level retouching and green-screen compositing.
Why this matters to you
You're probably not a big studio, so don't let "4K" and "50 references" dazzle you. Run the math: five 30-second product videos a week, Wan3.0 costs about 9,360 yuan a year in generation, Seedance 2.5 about 11k to 16k. The difference buys a decent camera.
Decide "do I need a finished clip, or do I need to show off" and the answer shows up.
Both vendors admit audio fidelity and text recognition still have rough edges, so don't expect broadcast-ready captioned output yet. Seedance 2.5's price is up about 50% over the last generation, so cost-sensitive teams should budget before diving in.
Wrap-up
This round of Chinese video models all reaching 30 seconds means "can it generate" is no longer the bar. "Affordable and usable" is. For related reading, the head-to-head on Kling, Runway and other video models shows where international and domestic tools each pull ahead.