Alibaba · Tongyi Lab
WAN 3.0
The current WAN line — prompt adherence, now with sound
WAN 3.0 is Alibaba Tongyi Lab’s video model, available on Ochra for text-to-video and image-to-video. It generates clips with synchronised audio and is the current WAN release, keeping the family’s prompt adherence while adding sound the earlier line did not produce.
Clip length: 4–30sResolution: 480p · 720p · 1080pAudio: Generated with the videoReleased: August 28, 2026
What is different about WAN 3.0
- Adds a synchronised audio track across the line.
- Keeps WAN’s defining strength: a long, specific instruction comes back as the shot you described rather than an average of it.
- Ships a Prime tier alongside the standard row for work that needs more finish.
Strengths and limits
Key Features
- Best-in-catalogue prompt adherence, which is why the family is chosen at all.
- Animates a supplied first frame faithfully, so a character stays the same character.
- Audio generated with the picture.
Where it falls down
- Still weak at text inside the frame. For a sign, a poster or a caption, use Nano Banana.
- Clip length is short next to the Seedance line; this family is chosen for fidelity to the prompt, not duration.
WAN 3.0 questions
3.0 is the current line and the better renderer; both generate audio. 2.5 is kept for price and for LoRA compatibility. If you have no reason to pin a version, use 3.0.