Back to models

Mischief 1

Video
tbc/mischief-1/image-to-videoView on fal.ai

Parameters

Dictation ready.

What happens in the video. Long, specific prompts steer WAN best.

http(s) URL of the condition image the video starts from, at most 25 MiB and 100 megapixels. A JPEG or PNG within the endpoint's byte budget (2.5 MB by default) renders as sent; anything larger, and any other format, is re-encoded to fit (a JPEG stays JPEG, anything else becomes lossless PNG), scaled down only as far as needed and never below the requested resolution; where that resolution stops the shrink, it is sent as a JPEG of slightly lower quality instead.

Clip length in seconds: 3 or 5. Rendered at 16 fps on the model's 4N+1 frame grid, so the clip runs slightly longer than asked: 3 s -> 49 frames (3.0625 s), 5 s -> 81 frames (5.0625 s). Billing follows the requested seconds, not the slightly longer rendered clip.

Result

Run the model to see the result here.