Background removal
Cut a photo down to its subject and take it away as a transparent PNG. There is no limit on how many you do, and the original never leaves your machine.
The original never leaves your machine
Both the photo you start from and the transparent PNG you end up with stay on your own device. That matters when you are working with pictures you cannot hand to anyone else.
There is no limit on how many you cut out, so you can keep trying until one comes out right.
Depth and pose on the node canvas
Depth maps and pose come from the Depth Map and Pose nodes on the node canvas, at 10 credits per image and, for video, 15 credits per second at 720p or 30 at 1080p. Pose picks up the body, feet, face and hands together.
Downloaded once
Only what you use is fetched — cutting out a background pulls only the files the cutout needs. Once fetched it comes from browser cache with no wait, and someone who never uses the feature never downloads anything.
Since the work is computed on your own device, use a computer
Carrying the result forward
A finished cutout is kept for you and can be fed back in as a reference image for the image models — generate a new background behind it, or place the same subject into a different scene.
There is also a route for applying this to video in the node canvas — those rates are on the pricing page
Frequently asked questions
Is there a limit on background removal?
No. You can keep swapping photos and cutting them out until one comes out right.
Is my photo uploaded anywhere?
The original you upload never leaves your machine, and the transparent PNG is produced in the same place.
Why is the first run slow?
That is when the files it needs are fetched. After that they come from browser cache with no wait, and someone who never uses the feature never downloads them.
Does it work on a phone?
Use a computer. The work is computed on your own device, so it is blocked on phones.
Can it extract anything besides the cutout?
Depth maps and pose come from the Depth Map and Pose nodes on the node canvas, at 10 credits per image and, for video, 15 credits per second at 720p or 30 at 1080p. Pose picks up the body, feet, face and hands together.
What format is the result?
A PNG, which is the format that carries a transparent background. Results are kept for you and can be fed back in as a reference image for the image models.