CLARU
// EXPLORE

A million-plus clips of the real world.

The shape of Claru's catalogue — where the footage was shot, what people are doing in it, how it was captured, and the structure that makes it ready to train on. Then play a representative selection yourself.

Clips in the catalogue
1M+Clips in the catalogueIndividual scene-cut clips, not source files.
Hours of video
20K+Hours of videoAcross long-form recordings and scene-cut footage.
Datasets
100+DatasetsDistinct collections, each with its own provenance.
Data categories
19Data categoriesFrom wearable video to speech, documents and evaluation.

Rounded figures for the catalogue as a whole. The clips you can play further down are a curated selection from it — open a folder to see the footage behind these numbers.

Environment

Where the footage was shot

Most of it is ordinary lived-in space — homes first, then the streets and green space around them, then the places people work and shop.

  • Homes — indoor~30%

    Kitchens · Living rooms · Bedrooms · Hallways & stairs

  • Homes — outdoor~25%

    Residential streets · Yards & patios · Garages & driveways

  • Streets, parks & nature~20%

    Parks & green space · Rural paths & farmland · Sidewalks

  • Workplaces, shops & venues~10%

    Retail stores · Workshops & job sites · Schools · Offices

  • Commercial streets & lots~10%

    Shopfronts · Outdoor construction · Parking lots

  • Vehicles & transit<1%

    Vehicle interiors · Stations & platforms

Shares are rounded to the nearest 5%. Every band opens into named sub-environments — 31 in all, from kitchens to parking lots — and you can open several of them further down this page.

Activity

What people are doing in it

Unscripted task footage rather than staged demonstrations: chores, errands, trades, driving and play, in the order you would expect to find them.

Broad coverage

  • Everyday household routines
  • Moving through streets & neighbourhoods
  • Yard work, grounds & the outdoors
  • Handling and manipulating objects

Substantial

  • Shopping, hospitality & front-of-house
  • Building, repair & job sites
  • Driving & in-vehicle
  • Screen & gameplay interaction

Specialist

  • Desk & office work
  • Teaching & classrooms
  • Care & clinical settings
  • Fitness & training
  • Transit hubs & platforms

Three tiers of emphasis across the catalogue — from the everyday routines that fill it to the specialist scenes we collect on request.

Capture

How it was captured

Six video-capture families, plus the non-video data the catalogue carries alongside them.

  • Head-mounted point of view

    Wearable capture of people doing real tasks, hands in frame.

  • Dashcam & in-vehicle

    Forward road scenes and cabin views from everyday driving.

  • Screen-captured gameplay

    Interactive 3D environments with synchronized input logs.

  • Internet-sourced video

    Curated open-web footage, clipped and captioned.

  • Cinematic & production

    Production footage across action and lifestyle.

  • Municipal & infrastructure

    Fleet, public-works and infrastructure monitoring capture.

Also in the catalogue

  • Product & e-commerce imagery
  • Computer-vision annotation
  • Video evaluation
  • Video understanding
  • Safety & moderation
  • Document understanding
  • Speech & audio
  • Multimodal instruction tuning
  • Code & math reasoning
  • Identity verification
  • Large-scale pretraining
  • Custom curations
  • Municipal & government

Signals that can travel with a clip

  • Scene captions
  • Synchronized IMU motion
  • Body pose tracks
  • Speech transcripts
  • Input & action logs
  • Per-clip technical specs

Each clip shows the signals it ships with on the clip itself, further down — so you can see the technical metadata before you ask for more.

Depth

Structured for training, not just stored

Footage is catalogued clip by clip — scene context to search by, technical metadata to filter by, and an environment vocabulary to browse by.

  • Scene contextClip by clip

    A written description of what is in frame, so footage is findable by content rather than by source file.

  • Technical metadataPer clip

    Duration, frame size, frame rate and codec carried on the clip, ready to filter against a training spec.

  • Environment structure6 bands · 31 folders

    Six setting bands resolving to named sub-environments, so you can browse by place and brief by place.

Open any folder below and the clip itself shows its description, its source specs and the signals it ships with.

Representative samples

Play some of it right now

A small curated selection from the catalogue above — not the catalogue itself. Pick the kind of footage you need, open a folder, and the clips play here. No login, no form.

Egocentric

First-person video data captured in crowded, public environments with diverse human interactions.

Portfolio context: approximately 118k hours of first-person footage across Claru's broader sourcing portfolio. Approximate, and the figure may overlap across sources.

Browse →

Dashcam & Traffic

Dashcam and in-vehicle traffic footage.

Portfolio context: approximately 1k hours of dashcam and in-vehicle footage across Claru's broader sourcing portfolio. Approximate, and the figure may overlap across sources.

Browse →

Internet Videos

Internet-scraped video collections

Portfolio context: approximately 1.5M hours in Claru's broader sourcing portfolio. Approximate, and the figure may overlap across sources.

Browse →

Gaming

Screen-captured gameplay footage from interactive 3D environments and game engines.

Portfolio context: approximately 25k hours of gameplay capture across Claru's broader sourcing portfolio. Approximate, and the figure may overlap across sources.

Browse →

Looking for something specific? We collect to spec.