Wan 3.0 vs Wan 2.7 Video

Specs side by side

Price per unit and per clip

Prices exclude taxes and are charged from your prepaid USD balance. Failed generations cost nothing.

Wan 3.0 is Alibaba's newer generation: 2-30 s clips from 480P to 4K, audio on by default and a much larger set of references. Wan 2.7 Video stays in the catalog for one reason: it can edit or extend an existing video, which Wan 3.0 does not do on Jantaris. For new shots I use Wan 3.0, because at 720P it is even slightly cheaper, and at 480P it costs less than half of the Wan 2.7 rate.

Key differences

• Length. Wan 3.0 makes 2-30 s in one pass, and with duration: -1 it picks the length itself (billed as 30 s). Wan 2.7 Video makes 2-15 s, or 2-10 s with a reference video or when editing.
• Resolution. Wan 3.0 offers 480P, 720P, 1080P, 2K and 4K. Wan 2.7 only 720P and 1080P.
• Audio. Wan 3.0 generates an audio track by default (generate_audio, no effect on price). The Wan 2.7 form has no audio toggle. An audio file in audio_url drives the motion of a still image, and at the vendor the reference mode can also assign voices to characters.
• Frames and references. Both accept a first and last frame. Wan 3.0 takes up to 10 images, 5 clips and 5 audio files as references; Wan 2.7 on Jantaris up to 4 images. Prompt: 20,000 characters in Wan 3.0, 5,000 in Wan 2.7.
• Editing. Only Wan 2.7 accepts a finished video to rework with a prompt (video_url) or to extend (video_urls). It also has negative_prompt, which Wan 3.0 lacks.
• Aspect ratios. Both have 16:9, 9:16, 1:1, 4:3 and 3:4. Wan 3.0 adds adaptive, which follows the input image.
• Price for 5 and 10 s. At 720P Wan 3.0 is about 3% cheaper, at 1080P about 1.17 times more expensive. A 5 s Wan 3.0 clip at 480P costs less than half of the same clip in Wan 2.7 at 720P. The 5 and 10 s amounts are in the table above.

When to pick Wan 3.0

• A single shot longer than 15 s, up to 30 s, without stitching.
• Cheap test clips at 480P from 2 seconds before you pay for the final resolution.
• A 2K or 4K shot for an ad or for reframing later.
• A scene built from many references: a character from several photos, motion from a clip, a voice from a recording.

When to pick Wan 2.7 Video

• Reworking part of an existing video with a prompt instead of generating from scratch.
• Extending a clip that ends too early.
• A still image animated by an audio track (image_url + audio_url).
• A shot where you need to rule something out in negative_prompt.

Cost in practice

Both bill per second by resolution. At 720P the rates are nearly equal; at 1080P Wan 2.7 is about 15% cheaper. Wan 3.0 at 2K costs about 1.7 times, and at 4K almost 2 times, what Wan 2.7 costs at 1080P. In both models the seconds of input videos are added to the bill, and in Wan 3.0 duration: -1 is billed as 30 s, so set the length explicitly for batches. Default rates: from $0.0777 per second and from $0.161 per second (1080P). Any combination can be priced in the AI cost calculator, and every model is on /pricing.

Wan 3.0 scored 7.75/10 in my two tests: good reflections on wet cobblestones, 2 errors in 12 words of a Polish sentence and music I did not ask for. I tested it through another provider's API; I have not tested Wan 2.7. Details in the best AI video generator in 2026, and Wan 3.0 against another model in Wan 3.0 vs Kling 3.0.

FAQ

Wan 3.0 or Wan 2.7: which one for new shots?

Wan 3.0. It has longer clips, higher resolutions, audio and more references, and at 720P it costs the same as Wan 2.7. Wan 2.7 makes sense for editing and extending an existing video.

Does Wan 2.7 generate audio?

The Jantaris form has no audio toggle for Wan 2.7. If you want an audio track in the file, pick Wan 3.0, where audio is on by default and does not raise the price.

How much do 30 seconds cost in Wan 3.0?

Thirty times the per-second rate at the chosen resolution. You see the amount before you start, in Studio and in estimated_cost, and a failed generation costs $0.

Vendor examples

Official examples from the model vendors, not our generations.