Vidlamp Stereo
Stereoscopic 3D from any video or image and its depth map: side by side, over/under, cross-eyed, red/cyan and red/blue anaglyphs. Z-buffered occlusions, holes filled from the background, steady depth across the clip, and a depth detail merge (Video Depth Anything + MoGe-2) for faces. Pure torch, no extra dependencies.
Vidlamp Stereo
Stereoscopic 3D for ComfyUI from any video (or image) and its depth map. Pure torch, no extra dependencies.
Nodes
Vidlamp Stereo 3D — images + depth (bright = near) → stereo frames.
output_format: side by side, side by side half, over/under, cross-eyed, red/cyan anaglyph (Dubois), red/blue anaglyph (half colour), left only, right only.divergence(% of the width, capped at 6),convergence(depth on the screen plane),balance,depth_gamma,edge_expand,smooth,normalize(one depth range for the whole clip by default),swap_eyes.- Occlusions with a z-buffer (the nearer pixel wins), uncovered areas filled from the background side.
Vidlamp Depth Detail Merge — base_depth (steady, e.g. Video Depth Anything) + detail_depth (sharp,
e.g. MoGe-2) → the large shapes of the first with the fine relief of the second, without its frame-to-frame flicker.
360 (equirectangular) — panorama_360 on Vidlamp Stereo 3D: the shift wraps around the left/right
borders, fades towards the poles and is zero at infinity (use side_by_side or over_under_full, the full-height
over/under). Vidlamp Project to 360 puts ordinary frames (a pose, depth or edges video) at the centre of a
black equirectangular canvas with a given horizontal field of view: the guide of a 360 video. Vidlamp Roll 360
turns a 360 video by half a turn (the seam goes to the middle to be repainted; the same node turns it back).
Typical graph
Load Video → Get Video Components → Video Depth Anything (depth, gray) → Vidlamp Stereo 3D → Create Video (with the original audio) → Save Video.