2.1 Sheffield Softworks Keyper

Isolate people from backgrounds

Keyper uses Machine Learning to extract people from their backgrounds without requiring greenscreen or prepared backdrops.

Keyper is not a replacement for traditional green screens used in special effects shots, rather it allows the user to apply filters and effects separately to the people or the backgrounds to create unique and engaging effects such as 3D still photos and the parallax effect.

Keyper – the people keyer – allows the user to apply filters and effects separately to the people and/or the backgrounds. People are not tracked between frames – each frame is analyzed on its own – and all people recognized by Apple’s Machine Learning are separated from the background.

What’s new in version 2?

Keyper 2.0 adds preprocessing capabilities to alter the image prior to recognition processing. These include brightness, saturation, contrast, sharpness, and the ability to load a LUT. These alterations are keyframeable as what may help in one frame may make recognition poorer in another. All of these adjustments do not alter the image that is output.

Also added are new matte processing capabilities, these include:

• Full Full matte. (same as Keyper 1.0)
• Outline Traces a line around the outside of the matte, controlled by Width and Edge.
• Inline Traces a line around the inside of matte, controlled by Width and Edge.
• Core Subtracts Inline from the matte to retain the Core, controlled by Width and Edge.
• Combo Combines Core and Inline. Useful for filling holes in the matte.

As well as the initial Keyper output offerings of Composite, Matte, and Matte over Background, Keyper 2.0 adds Colored Matte – a colored matte with Alpha.

Finally, Keyper 2.0 adds Patch which you can use to patch a hole in your matte, or to block out something or someone you don’t want to be included, or you can use it to only include the person covered by the patch.

New in Version 2.1

The new Subject detector (for macOS 14 Sonoma+) will identify human and non-human subjects and can sometimes do a better job of selecting people.