Catalog Custom orders Languages Quality Request a dataset
REC — LIVE COLLECTION IN 63 LANGUAGES

Raw material
for machine
intelligence

Grey Global records, sources, and labels audio, image, and video data for AI teams — off the shelf or built to your spec, in any language your model needs to learn.

AR-EG · 40H · VERIFIED
JA · SPONTANEOUS
SW · NEW
PT-BR · TAGGED
LANGUAGES 63 HOURS RECORDED 210,000+ IMAGES TAGGED 4.2M ACTIVE COUNTRIES 41 CONTRIBUTORS 38,000+
Modalities

Three kinds of raw material

Every dataset starts as audio, image, or video — captured to your protocol, then reviewed by a native-language team before it ships.

REC-Aid: audio

Audio / Voice

Scripted and spontaneous speech, call-center dialogue, wake-word repetitions, accented and code-switched recordings.

ASRTTSWake-wordDiarization
REC-Iid: image

Image

Photographs, scene captures, and document scans, shot or sourced to your bounding-box or segmentation spec.

DetectionOCRSegmentationClassification
REC-Vid: video

Video

Scripted scenarios, gesture and action clips, ambient footage, and multi-camera capture for motion models.

ActionGestureMulti-angleAmbient
Ready-made

From the catalog

A running set of pre-collected, pre-reviewed datasets available now. Every entry is a real record — request access and we confirm licensing terms with you directly.

DS-0142

Conversational Arabic (Egyptian) Speech

40 HRS · 16KHZ WAVTRANSCRIBED + VERIFIED
DS-0098

Everyday Kitchen Objects

12,000 IMAGES · COCO FORMATBOUNDING BOXES
DS-0231

Retail Storefront Video

300 CLIPS · 1080PACTION-LABELED
DS-0176

Swahili Read & Spontaneous Speech

65 HRS · WAVVERIFIED TRANSCRIPTS
DS-0304

Hand Gesture Video Set

8,500 CLIPS · MP4LANDMARK-ANNOTATED
DS-0059

Multilingual Form Scans

25,000 PAGES · 14 LANGUAGESOCR-READY
Custom orders

Built to your spec, in five steps

Need something the catalog doesn't have yet? Custom collection follows the same pipeline every time, in any language, from a single locale to a global rollout.

STEP / 01

Spec

You define language, modality, volume, and format. We scope contributors and timeline.

STEP / 02

Recruit

We source native contributors in the target locale, screened and consented.

STEP / 03

Capture

Recording, photography, or filming happens in-field or in-studio, to your protocol.

STEP / 04

Review

Native-language QA, transcription, and labeling pass before anything ships.

STEP / 05

Deliver

Data lands in your format and schema, with an audit trail for every file.

Global reach

Wherever your users speak, we can record it

63 languages and locales are live today. New ones are onboarded per request, typically within two to three weeks.

Don't see your target locale? Ask — sourcing a new one is a standard custom order.
Quality & compliance

Every file is traceable

Trust matters more than volume. Here's what stands behind every dataset that leaves Grey Global.

Consent-first sourcing

Every contributor signs a locale-appropriate consent and usage agreement before recording starts.

Native review

Transcripts and labels are checked by reviewers fluent in the source language, never machine-translated.

Secure handling

Encrypted transfer and storage, with access limited to your named project team.

Full audit trail

Every file is traceable to a contributor ID, capture session, and QA pass.

Used for

Built for the models you're training

Speech recognition Computer vision Conversational AI & LLMs Document intelligence Autonomous systems Biometrics
Certifications

Standards we operate by

Independently audited, renewed on schedule. Documentation available on request under NDA.

ISO/IEC 27001

Information security management

GDPR Compliant

EU data protection & processing

SOC 2 Type II

Security, availability & confidentiality

ISO 9001:2015

Quality management systems

Tell us what your model needs to learn.

Send the shape of the project — modality, language, volume — and we'll come back with scope and timeline.

EMAIL — ceo@grey-global.com RESPONSE TIME — within one business day
No commitment — this starts a scoping conversation.
Request received. We'll be in touch within one business day.
Something went wrong. Please try again or email us directly.