The latest Wanxiang 2.1 720P model as the core, directly outputs high-quality 1080P video in 6 seconds, allowing separate control of character movements and camera direction.

Using prompt-based reverse deduction (character consistency is more accurate, movements are smoother), check the instructions below to configure and directly output videos longer than 10 seconds.

Frame setting instructions: 123456 seconds


Note: 161 frames use processor region adjustments to limit the original generation width and height to 512. If the original limit is set to 768, the highest output is 73 frames, which is a 6-second video. Higher than this will result in an error, which can be resolved by increasing the block swap value to 40. It is recommended to output 5-second videos, which is 61 frames, for optimal results.

Final output width and height: Regardless of the size of the input image, it will ultimately be center-cropped to this size.


WanVideo BlockSwap: For the same number of iteration steps (10), the larger the block swap value, the higher the resolution that can be generated, but the slower the processing time. Normally, a 6-second video with a 768 original image limit set to 10 can run. If set to 1024 or even 1280, it is recommended to increase the block swap value to 30 or 40 to avoid memory overflow. You can also increase the duration for the same resolution, which requires increasing the block swap size.

High-definition quality settings: Original image limit 768, output arbitrary, iteration 30, block swap 10, maximum video duration 6 seconds (73 frames), generation time (10 minutes)

High-quality settings: Original image limit 720, output arbitrary, iteration 20, block swap 10, maximum video duration 6 seconds (73 frames), generation time (8 minutes), can try 7 seconds (85 frames)


Medium-quality settings: Original image limit 512, output arbitrary, iteration 10, block swap 10, maximum video duration 6 seconds (73 frames), generation time (5 minutes), can output 10 seconds (121 frames), maximum 161 frames, 13.5 seconds (tested)

Low-quality settings: Original image limit 384, output arbitrary, iteration 10, block swap 10, maximum video duration 6 seconds (73 frames), generation time (3 minutes), can try longer durations