Wan 2.5
Alibaba's video model with native audio-visual sync — 10s 1080p clips from text or image.
About
Wan 2.5 is Alibaba's AI video generation model, released September 2025 on Alibaba Cloud's DashScope platform. Its headline feature is native audio-visual synchronization: a single prompt produces video with synchronized voices, sound effects, and music in one pass — no separate lip-sync step. It supports text-to-video and image-to-video at 480p, 720p, and 1080p, up to 10 seconds per clip at 24fps, and debuted at #5 on the text-to-video leaderboard. Unlike earlier Wan releases, 2.5 is closed-weight and API-only.
The honest picture: it's the affordable Veo 3 alternative at around $0.15/second, with genuinely good multilingual prompt handling including Chinese. But the closed-weight pivot means no self-hosting or fine-tuning (unlike Wan 2.1/2.2), and 10 seconds is still short-form territory. Reviewers rank it below Veo 3 and Sora 2 Pro for high-end client deliverables.
What it can do
- Text-to-video and image-to-video, up to 10s at 1080p/24fps
- Native audio generation synced to visuals in one pass
- Custom voice input or model-generated voices with lip-sync
- Built-in prompt expansion and optimization
- 6 aspect/size options for social and widescreen
- Priced around $0.15/second via API
Setup
Access via the Alibaba Cloud DashScope API or third-party platforms (WaveSpeed, Higgsfield, Xole AI) that host the model.
Prerequisites
A DashScope API key or an account on a hosting platform; prompts work in English and Chinese.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 46; Trending score 78/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.