← Back to all generators

andreasjansson/wan-1.3b-inpaint

Inpainting and video2video experiments with Wan 2.1

Capabilities

Negative PromptSeed

Cost

Community model (estimated from hardware time)

Input Parameters

input_videorequiredstring

Original video to be inpainted

promptrequiredstring

Prompt for inpainting the masked area

expand_maskinteger

Expand the mask by a number of pixels

Default: 10min: 0, max: 100
frames_per_secondinteger

Output video FPS

Default: 16min: 5, max: 30
guide_scalenumber

Guidance scale for prompt adherence

Default: 5min: 1, max: 15
inpaint_fixup_stepsinteger

Number of steps for final inpaint fixup. Ignored when in video-to-video mode (when mask_video is empty)

Default: 0min: 0, max: 10
keep_aspect_ratioboolean

Keep the aspect ratio of the input video. This will degrade the quality of the inpainting.

Default: false
mask_videostring

Mask video (white areas will be inpainted). Leave blank for video-to-video

negative_promptstring

Negative prompt

Default: ""
sampling_stepsinteger

Number of sampling steps

Default: 50min: 20, max: 100
seedinteger

Random seed. Leave blank for random

Default: -1
strengthnumber

Strength of inpainting effect, 1.0 is full regeneration

Default: 0.9min: 0, max: 1
Version: 7abfdb3370abUpdated: 7/25/2026498 runs