From raw footage to a reusable transparent layer.
Four automated stages follow the subject through motion, preserve partial transparency at the edge, and encode the result into a format that keeps its alpha channel.
Detect
The AI finds your subject automatically, so you do not need to draw an outline before processing starts.
Automatic subject detection, person-optimised with an object fallback
Track
The AI follows your subject across frames to reduce flicker and keep edges stable as the video moves. That is the difference between video matting and running an image cutout on every frame separately.
Per-frame propagation with a memory of earlier frames
Solve
Per-pixel alpha preserves partial transparency around hair, motion blur, and semi-transparent fabric instead of forcing every edge into a hard cutout.
Continuous alpha, 0 to 1, not a hard cutout
Encode
BGRemove exports the result as transparent WebM, preserving the alpha channel for editing and compositing. It plays on the web and imports into an editor as a layer with its transparency intact.
Alpha preserved through encode, no flatten step
Why running an image tool on every frame does not work
Most free background removers are image models in a loop. On a still photo they can be excellent. On video they produce the artefact everyone recognises immediately, even if they cannot name it.
Frame-by-frame image cutout
Each frame is solved with no knowledge of the last one
Edges shimmer and hair appears to breathe between frames
A subject briefly occluded is often lost entirely
Lower processing cost, but prone to visible frame-to-frame inconsistency
Temporal video matting
Each frame inherits context from the frames before it
Edges stay put, so composited footage sits still
The subject is easier to keep track of when an object briefly passes in front
What BGRemove runs
Output formats
Transparent WebM using VP9 with alpha. Output preserves the source video’s dimensions.
WebM / VP9
Transparency stored in the file itself. Whether it displays depends on the player or editor reading VP9 alpha.
A simple workflow from upload to download
Upload your clip, follow the processing status, and download a transparent result. The current workflow prioritizes fast automatic matting and editor-ready output, with room for meaningful refinement tools as the product develops.
- Typical 30s clip
- Under 2 min
- Processing region
- EU
Keep the subject. Remove the setup.
Upload real footage and download a transparent result ready to become the next layer in your edit.
3 free videos per 24-hour allowance window. No watermark. No card.