media-depth

Estimate monocular depth maps from images or videos using Depth-Anything v2 and MiDaS.

15|4|Updated Apr 18, 2026

Search Tags:#parallax #vfx #stereo #depth-estimation #monocular-depth #depth-anything #midas

Ready to run

npx skills add https://github.com/damionrashford/media-os --skill media-depth

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: media-depth
Source: https://github.com/damionrashford/media-os/tree/main/skills/media-depth
Command: npx skills add https://github.com/damionrashford/media-os --skill media-depth

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires transformers, torch, opencv-python, numpy, pillow, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This skill provides reliable monocular depth estimation from a single image or video, enabling depth-driven compositing, 2.5D parallax, and stereo generation without specialized hardware.

Core Features & Use Cases

Depth maps for VFX pipelines: generate per-pixel depth to drive Blender, Unreal, Nuke, or compositing passes.
2.5D parallax & stereo: create side-by-side or left/right views from flat footage or photos for immersive visuals.
Model options & safety: uses Depth-Anything v2 and MiDaS with commercial-safe licenses for broad applicability.

Quick Start

Run the depth.py image subcommand with an input image to generate a depth map.

Frequently Asked Questions about media-depth

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I generate a monocular depth map from a single image?▼

To generate a monocular depth map from a single image, run the depth.py image subcommand with your input image. The skill uses Depth-Anything v2 and MiDaS models to estimate per-pixel depth from flat footage or photos.

Can I use these depth maps for compositing in Blender, Unreal, and Nuke?▼

Yes, you can use these depth maps for compositing in Blender, Unreal, and Nuke. The skill generates per-pixel depth data specifically to drive depth-based compositing passes and 2.5D parallax within VFX pipelines.

Do I need specialized hardware for monocular depth estimation?▼

No, you do not need specialized hardware for monocular depth estimation. The skill estimates depth from a single image or video using ML models like Depth-Anything v2 and MiDaS, requiring only the standard Python ML stack including torch and transformers.

What's the best way to create side-by-side stereo views from flat footage?▼

The best way to create side-by-side stereo views from flat footage is using monocular depth estimation to generate left and right views. This skill uses Depth-Anything v2 to produce depth maps that enable 2.5D parallax and stereo generation.

Are the Depth-Anything v2 and MiDaS models safe for commercial VFX projects?▼

Yes, Depth-Anything v2 and MiDaS are safe for commercial VFX projects. The models use open-source licenses, specifically Apache 2.0 for Depth-Anything v2 and MIT for MiDaS, ensuring broad commercial applicability without licensing barriers.