AI-Powered Pipeline

How Teleport AI Works

Every piece of 360° content passes through a 7-step AI pipeline that understands scenes, scores quality, maps depth, and sets pricing — fully automated.

Step 1

Content Upload

When new 360° content is uploaded, an event fires to start the async AI pipeline. Processing runs entirely in the background — creators never wait.

Async Event Queue

upload complete → pipeline triggered → background processing

Non-blockingFully automated
Step 2

Visual Understanding

The content is processed through a vision model that generates a high-dimensional vector embedding — a numerical fingerprint that captures what the scene looks like, enabling semantic search and similarity matching.

Vision Embedding Model

360° image → 512-dim vector → indexed for fast search

Visual embeddingSemantic searchSimilarity matching
Step 3

Parallel Processing

Three AI models run simultaneously, each extracting different intelligence from the content.

Runs in parallel
Step 3a

Smart Tagging

Zero-shot classification against 125+ candidate labels. No training needed — tags are inferred directly from the visual content.

Vision ClassifierAI Tags
Step 3b

Scene Description

A multimodal language model analyzes the 360° image and writes a natural-language description of the space, objects, and atmosphere.

Multimodal LLMAI Description
Step 3c

Auto-Categorization

AI matches the content against existing marketplace categories, returning the best-fit category with a confidence score.

Language ModelAI Category
Step 4

Quality Scoring

A vision model evaluates the content across 4 dimensions — resolution, stitching, lighting, and composition — each scored 0-25 for a total 0-100 composite score.

Vision Quality Model

4 dimensions × 0-25 pts = 0-100 total → Gold / Silver / Bronze

Quality scoreTier badgeScore history
Step 5

Depth Estimation

A monocular depth model generates a depth map from the 360° image, revealing the spatial structure of the scene. The result is stored and available as a toggleable overlay.

Depth Estimation Model

360° image → depth map → stored on CDN → toggle overlay

Depth mapSpatial dataVisual overlay
Step 6

Smart Pricing

A pricing engine combines quality score, content uniqueness (how visually distinct the scene is), and estimated market demand to suggest an optimal licensing price.

Pricing Engine

base × quality × uniqueness × demand = suggested price

Suggested priceFactor breakdownMarket context
Step 7

Marketplace Ready

All AI outputs are stored and the content is marked as fully processed. It's now discoverable via semantic search, browseable by category, and ready for licensing.

SearchableTagged & describedQuality scoredDepth mappedPrice suggestedCategorized

Ready to see it in action?

Upload your 360° content and watch the AI pipeline process it, or browse the marketplace to explore what's already been analyzed.