Depth Anything

Vendor: Depth Anything · Official docs: Video Depth Anything on GitHub

Video Depth Anything estimates temporally consistent depth for every frame of a video and renders the result as a new clip — grayscale by default, or a color map (turbo, inferno, magma, viridis), optionally side by side with the source. Pick the VDA-Small tier for speed or VDA-Large (the default) for the most accurate depth. It is a video-to-video model: the video is the only required input and there is no prompt.

Mode: video · Models: 1

Models

id	Name	Input type
depth-anything-video	Video Depth Anything	v2v

CLI

# grayscale depth map at the source resolution
gen-ai generate -m depth-anything-video --video ./clip.mp4

# color-mapped depth, side by side with the source, at 720p
gen-ai generate -m depth-anything-video --video ./clip.mp4 \
  --colormap turbo --side-by-side -r 720p

MCP

{ "name": "picsart_generate",
  "arguments": {
    "model": "depth-anything-video",
    "videoUrl": "https://example.com/clip.mp4",
    "colormap": "turbo",
    "sideBySide": true
  } }

Parameters

Full parameter surface for every model, sourced from gen-ai models info <id> --json. CLI flags show the primary short form; the canonical --kebab-case long form always works too.

depth-anything-video — Video Depth Anything

Try depth-anything-video in Playground ↗

Input type: v2v

Param	CLI flag	Type	Values
videoUrl	--video	file	required video
model	--model-version	enum	VDA-Small · VDA-Base · VDA-Large (default VDA-Large)
colormap	--colormap	enum	grayscale · turbo · inferno · magma · viridis (default grayscale)
resolution	-r	enum	auto · 360p · 480p · 720p · 1080p (default auto)
sideBySide	--side-by-side	boolean	true · false (default false)