Wan 2.2 is Alibaba's frame-controlled workhorse. It's simpler than Wan 3.0, charges by clip instead of by second, and skips audio to keep generation fast. Use it for quick text-to-video when you're building b-roll or silent video sequences.

Why creators choose Wan 2.2

Wan 2.2 trades some of Wan 3.0's capabilities for simplicity and predictable cost. If you're generating short clips and adding sound in post, Wan 2.2 keeps the workflow lean. The frame-controlled architecture means you get consistent length and fixed per-clip pricing, which matters when you're building dozens of clips into a longer piece.

Predictable per-generation pricing

Wan 2.2 charges per clip, not per second. You know the cost upfront. No length multiplier, but resolution changes the per-generation cost. That clarity speeds up budgeting for projects with many clips. When you're cutting a montage with 20 or 50 clips, fixed pricing beats per-second models that scale with your creative choices.

Lean and fast

Without audio generation overhead, Wan 2.2 focuses on video quality and generation speed. It's purpose-built for creators who handle sound in the edit, either adding voiceover, music, or sound effects in post. This keeps the generation time down and the workflow simple, which helps when you're iterating on b-roll.

Multiple aspect ratios for different cuts

Wan 2.2 supports 16:9, 9:16, and 1:1. Shoot one brief for widescreen and vertical simultaneously. Repurposing the same clip for mobile and desktop doesn't require running separate generations.

Mixture-of-experts architecture for cleaner visuals

Wan 2.2 uses two expert networks, one for noisy inputs and one for clear ones. The model sorts inputs intelligently and applies the right pathway, resulting in steadier visuals and fewer artifacts. This shows up as more stable motion and cleaner edges, especially in fast pans or complex scenes.

How to use Wan 2.2 in Craze

Wan 2.2 is straightforward: describe the shot, Craze generates it, you build around it.

1. Start a project and describe the shot you want. For Wan 2.2, be clear about the action, setting, and mood.
2. Pick Wan 2.2 in the Models panel, or name it in chat. Craze shows the cost before planning.
3. Craze generates the clip at your chosen resolution and aspect ratio.
4. Add the clip to the timeline and layer sound, voiceover, music, or effects on top.

Be specific in your prompt. "A slow dolly toward a coffee cup on a wooden table" gives Wan 2.2 clear direction, while "something cinematic" leaves the model guessing.

Wan 2.2 pricing

You pay $0.211 per clip. The cost depends on resolution: 480p costs less than 720p. Wan 2.2 charges per generation, with no duration choice in the Craze calculator.

This fixed-cost model is convenient for projects where you're generating dozens of short clips and handling sound in post. The cost calculator on this page shows both resolutions so you can budget for your montage upfront. If you're building a 30-second edit from 20 short clips, you can estimate the total cost of all generations in advance, which simplifies project planning and client quotes.

Who Wan 2.2 is for

B-roll and montage sequences

Build a montage of quick clips with text prompts, then layer music and sound effects. Wan 2.2 gives you the raw visual sequences fast. The per-clip pricing makes it affordable to generate multiple variations and choose the one that works for your cut.

Silent video sequences for voiceover

If you're recording voiceover or adding a music bed, Wan 2.2 keeps the video generation quick and cost-predictable. Edit the voiceover first, then generate b-roll to match your read, and layer them together.

Social media snippets without commentary

Short, silent video clips work well for social content when paired with text overlays or trending audio added in post. Wan 2.2's fixed pricing scales well when you're generating lots of clips for a feed.

Wan 2.2 vs other models

Wan 3.0 is longer (up to 30 seconds), generates synchronized audio, and supports reference images or first and last frames. Choose Wan 3.0 when you need length, audio sync, or character consistency with references. For even longer clips with more reference control, Seedance 2.5 runs 30 seconds and takes up to 30 reference images.