Remove People & Objects from ANY Video with AI — VOID API Tutorial (3 Engines Compared)
Remove unwanted people, objects, photobombers, logos, watermarks, and even shadows from any video — automatically, with one API call.
In this tutorial we walk through PixelAPI's VOID Object Removal endpoint end to end on a real travel clip from Colva Beach, Goa.
What you'll see:
One JSON request that removes two strangers from a 16-second 4K clip
The 4-step pipeline: SAM2 mask, light dilation, model fill, full-resolution stitch
All THREE engines side by side on the same shot:
• DiffuEraser (default) — cleanest AI-generated fill at 1440x1080, best for most jobs
• ProPainter — flow-based, sharper texture, faster, best when geometry behind the subject is simple
• Netflix VOID — physics-aware diffusion (CogVideoX backbone), best on tricky shadow and shading reconstruction
Use cases for video editors and creators:
Wedding videographers wiping out photobombers from ceremony footage
Travel vloggers cleaning crowded landmarks (Taj Mahal, Eiffel Tower, beach paths)
Real estate walkthroughs removing staging crew, cables, ladders
Sports / fitness creators wiping branded logos for re-licensing
Stock footage prep — remove people to make plates royalty-safe
Pricing:
80 credits per generation (about $0.80 at the starter tier)
Same price across all three engines — pick what fits the shot
4K input supported, 1440x1080 output on the default engine
API endpoint:
POST https://api.pixelapi.dev/v1/video/remove-o...
Inputs:
video_url (mp4)
object_prompt ("the two men walking on the path")
engine (optional: diffueraser / propainter / netflix_void)
Output:
A clean mp4 with the subject and its shadow gone, downloadable from /dl/
Get started free at https://pixelapi.dev — every signup gets free credits to try VOID.
Docs: https://pixelapi.dev/docs
Pricing: https://pixelapi.dev/pricing
Chapters:
0:00 What VOID does
0:20 The Colva Beach example
0:40 Step 1 — auto-segment with SAM2
1:05 Step 2 — light dilation for halo-free edges
1:30 Step 3 — model-fills the void
2:00 Step 4 — full-resolution stitch back
2:30 Honest comparison: where the haze still shows
3:15 Three engines side by side
4:30 Pricing and how to get started
Built with: SAM2 (segmentation), DiffuEraser, ProPainter, CogVideoX-Fun-V1.5-5b-InP (Netflix VOID).
Subscribe for more PixelAPI tutorials — we ship a new endpoint every week.