Segment the person
The foreground video is analyzed to isolate the body shape from the frame.
Turn a person mask into animated grid cells. The subject appears only inside the squares touched by their body, while the rest of the frame keeps your background video.
This first version accepts public MP4 links. Uploads will come later; for now, paste a person clip and a background clip that Replicate can fetch.
The creative trick is not a soft cutout. It is a hard grid decision: if the person touches a square, the whole square becomes foreground.
The foreground video is analyzed to isolate the body shape from the frame.
Touched pixels are converted into whole grid squares, creating a crisp tile silhouette.
Person cells reveal the foreground clip while every other cell keeps the background.
Grid lines and settings are applied consistently across the sequence.
This page is the first effect template: a public landing page, a runnable form, an app API route, a Replicate model, and reusable launch rules for the next presets.
Make a person cut through background footage with a tile-by-tile reveal that reads instantly on mobile.
Place a performer over texture, ocean, city, or abstract plates without using a normal soft mask.
Reveal looks, silhouettes, and walking shots in a graphic grid system that feels editorial.
Prototype a recognizable motion treatment before committing to a full post-production pass.
Expose the preset as a repeatable video effect inside your own app, queue, or creative automation flow.
Turn one visual rule into reusable settings for different grid densities and mask expansion styles.
9 columns, 12 rows, slightly expanded mask for a clean body silhouette.
More cells for a detailed reveal that keeps the person shape more readable.
Larger squares and more mask expansion for a louder editorial look.
The public route sends jobs to the server-side endpoint, which keeps provider tokens private and returns a prediction id for polling.
POST /api/effects/grid-reveal
{
"personVideoUrl": "https://...",
"backgroundVideoUrl": "https://...",
"columns": 9,
"rows": 12
}It is a video effect that segments a person, maps the mask to a grid, and reveals the person video only in grid cells touched by the body. The remaining cells show the background video.
This first public version accepts hosted MP4 URLs. A direct upload flow is the next layer to add once storage is wired into the app.
Yes. The page calls the BLXSSXM app endpoint at /api/effects/grid-reveal, which proxies the Replicate model and returns a prediction id and output URL.
The effect depends on human segmentation. The mask expansion setting helps grab more grid cells around the body, but very low contrast or unusual poses can still reduce accuracy.
It is an early preset endpoint. It is useful for testing, demos, and internal workflows, but upload handling, GPU scaling, and queue UX should be hardened before high-volume use.