Lab · 2026
ZPass
A desktop tool that converts any video into a depth map sequence. Built for VFX artists and compositors who need clean depth passes without cloud services, subscriptions, or API keys.
OriginalDepth
Features
01
Offline Inference
Runs entirely on your machine. The model downloads once and lives in local cache. No internet required after that.
02
Live Adjustment
Near/Far clip, Gamma, Blur radius, Invert, Colormap — all applied in real time on the cached depth data. No re-inference.
03
Global Normalization
Depth values are normalized across the full clip before export — eliminating inter-frame flickering entirely.
04
Non-Destructive
Analyze once, export as many variants as needed. Switch colormaps, adjust gamma, re-export — no waiting.
05
Frame Scrubber
Inspect any frame with a single drag before committing to an export. Depth and original stay in sync.
06
Three Preview Modes
Depth only, split view, or interactive wipe — drag a bar to compare original and depth on the same frame.
Preview Modes
Depth
Full-frame depth map with colormap applied. All adjustment parameters update the preview in real time.
Split
Original video on top, depth map below. Both scrub together — precise frame-level comparison.
Wipe
Drag a vertical bar to reveal depth vs original. The same compositing tool you know from Resolve and Nuke.
Parameters
Near clip
0.0 – 1.0
Clips the near boundary of the depth range. Push inward to remove very close elements.
Far clip
0.0 – 1.0
Clips the far boundary. Pull inward to focus the range on the subject.
Gamma
0.1 – 4.0
Redistributes midtone values. Below 1.0 pushes depth toward near, above 1.0 toward far.
Blur radius
0 – 20
Gaussian blur on the depth output. Smooths noisy edges without affecting the inference data.
Invert
on / off
Flips the near/far mapping — white becomes far, black becomes near, or vice versa.
Colormap
6 options
Grayscale · Inferno · Turbo · Viridis · Magma · Jet. Applied at export, not during inference.
Stack
Python 3.11Depth Anything V2PyTorchHuggingFace TransformersOpenCVCustomTkinterPillowNumPy