A set of separate points placed in 3D space. Motes places each point where its pixel would sit in depth, so a flat picture gains real volume when the camera moves.

Smart depth runs Depth Anything V2, an open neural network, inside your browser. It estimates how far away each part of the picture is. Your first image downloads it once, about 27 MB, and your browser keeps it for next time. Shape and Brightness work instantly without it.

No. Reading the image, working out the depth and rendering the video all happen on your device. Only the model file travels, from Hugging Face to you.

Yes. Add an MP4 or MOV and every frame becomes points, with depth taken from brightness so it follows the motion. Sound isn’t included in the export, so add it back in your editor. Motes uses up to 60 seconds on a computer and 20 on a phone.

MP4 at 9:16, 4:5, 1:1 or 16:9, up to 4K on a capable computer, at 24, 25, 30, 50 or 60 fps, from 2 to 20 seconds. Every frame is rendered one by one, so none get dropped however heavy the scene is.

Yes. Choose Transparent and Motes makes a ZIP of PNG frames with an alpha channel, which After Effects imports as a sequence. MP4 can’t store transparency. If you prefer one video file, keep it on black and use a Screen blend mode.

Yes. What you make with Motes is yours, as long as you have the rights to the image you put in.

Current Chrome, Edge, Safari and Firefox, on computers and phones. Motes needs WebGL 2. Phones get fewer points automatically so the preview stays smooth.