There are three keying modes, „Background“, „Color“ and “Automask”.
In this mode you provide the plugin with a video or image of background+subject and a still image of the background alone. Best, during recording you make sure to capture a frame without the subject. Lightning between both clips should be as similar as possible and the colors of the foreground subject shouldn’t be present in the background layer. Both layers don’t have to align 100% percent. If they contain enough visual clues (like the wall pattern in the example above, text boards or other features of high contrast) they can be perspectively aligned by the plugin.
This is the layer with the clean background plate as described above.
If your foreground+background video and your background plate don’t match 100% (e.g. the foreground+background video was shot handheld) switch this on and both layers will be perspectively aligned. You can always check the result of the alignment by setting „Output Mode“ to „Warped Background“. If your footage already aligns very well (because it was shot from a tripod), leave this off.
Maximum Matchpoints (only with Match Perspective)
Set maximum amount of features to find for perspective matching. Increase this value if perspective alignment doesn’t produce stable result.
Use Percentage of Matches (only with Match Perspective)
The higher you set this value, the more percentage of the features found are used for perspective matching. E.g. setting this to 10% will use the best 10% of all features.
The AI engine used internally. „Low“ can be alot faster than „Ultra“, but „Ultra“ gives the best result.
Refinement of the edges. Set to a higher value at the cost of performance.
This is a simple AI based color keyer, in which you only set a color. Color mode can be used if the background is of uniform color and doesn’t require an background layer.
Output Mode, Engine & Refine Threshold as above
The color to key.
Choose between several models, finetuned for persons and generic object. New: Generic (2022)
Choose between „Result“, „Matte“, „Warped Background“ or „Matchpoints“.
Backend & Performance
Hardware Acceleration (on GPU version of the plugin only)
Run calculations on the GPU. This will give massive speedups compared to CPU mode.
Compute with reduced precision if possible. This can save up to half of the memory and give you some speedups.
Optimize for low VRAM
If you have not that much graphics card memory enabling this will try to optimize memory consumption.
Samples (not available for all settings)
The number of ai samples to calculate. This will improve the models accuracy.
Computation Tiles (not available for all settings)
Split the computation into several tiles. This can help if you run out of memory.
You get a message „Not enough features to match.“ This can happen if you use the „Match Perspective“ feature, but your foreground+background and background video don’t contain enough visual clues to match the perspective. Set the „Use Percentage Of Matches“ to a higher value. Experimenting with this value may be required, since a higher value will also incorporate features of lower quality into the matching process. Best solution is to make sure during the recording, that your footage contains enough visual clues. If you are in a studio setup, you could place tracking markers in the corners of your frame (they shouldn’t intersect with the foreground subject!).