Wan AI 2.2 A14B: Text & Image to Video Online
Bottom line
Alibaba's open MoE video model, tuned for precise 1–5 second clips in five aspect ratios. Start from a prompt or a still image in the Text to Video tool; VdoBloom shows the exact credit cost before each Wan AI 2.2 A14B run, with no separate Wan AI account.
Wan AI 2.2 A14B is the architecture pick in the Wan lineup. The A14B name refers to its Mixture-of-Experts design — roughly 14 billion parameters active per generation step — which Alibaba introduced with the openly released Wan 2.2 generation. In practice, that means a model that routes different parts of the job to specialized experts rather than pushing everything through one dense network.
What you get on VdoBloom is deliberately compact: clips from 1 to 5 seconds, selectable in one-second steps, rendered at 720p across five aspect ratios (16:9, 9:16, 1:1, 4:3, and 3:4). It handles both text-to-video and image-to-video from the same page, so one model covers a prompt-only idea and an animated still alike.
The short ceiling is a feature rather than a limitation. One- and two-second clips are exactly what transitions, loops, and motion inserts need, and they are quick to iterate on. Runs are charged in VdoBloom credits, with the price for your settings displayed before each generation.
What creators use Wan AI 2.2 A14B for
- Generate 1–2 second motion inserts and whip transitions for edits without rendering length you would only trim away.
- Make looping 1:1 or 9:16 animations for profile headers, stickers, and animated backgrounds.
- Test a prompt as a 1-second 720p clip before committing it to a longer, higher-resolution model.
How to use Wan AI 2.2 A14B on VdoBloom
- 1Open the Text to Video tool and select Wan AI 2.2 A14B from the model picker.
- 2Write your prompt — or upload a starting image for image-to-video.
- 3Choose your duration (1–5s), quality (720p), aspect ratio, review the credit cost, and generate.
Frequently asked questions
What does the A14B in the name mean?
It refers to the model's Mixture-of-Experts architecture, with about 14 billion parameters active per step. Alibaba introduced this MoE design with the openly released Wan 2.2 generation, routing work across specialized experts instead of one dense network.
Why does Wan 2.2 A14B only generate 1 to 5 seconds?
It is built for short-clip precision: every length from 1 to 5 seconds is selectable in one-second steps. That range suits transitions, loops, inserts, and quick prompt tests better than long single takes.
Does it handle both text-to-video and image-to-video?
Yes, both modes run from the same model page. Either way, output is 720p in your choice of 16:9, 9:16, 1:1, 4:3, or 3:4.
What content rules apply to Wan 2.2 A14B?
The model is labeled HOT on VdoBloom, meaning its filtering is more permissive and flexible than typical mainstream models. Platform-level rules still apply, but it is one of the less restrictive options in the catalog.