Video 104 scenes
Show the world. Describe what's happening.
Upload a clip and describe what's happening on screen. Reviewers add scene labels and temporal boxes.
Street food walkthroughsHow-to clipsSports highlights
Public square · open
Where the world labels
and AI learns.
A public square for datasets. Upload a clip and say what's in it. Tag a plant on your walk. Read a proverb aloud. Small human contributions become structured data anyone can train on.
WKALJ
4,892 contributors building public datasets right now - data formats
- 4
- public datasets
- 12.4k
- reviewer agreement
- 94%
- task routing
- 24/7
Formats
Four formats, one operating model.
Each format gets a purpose-built surface. Project setup, assignment, review, and exports stay consistent across the platform.
Image918 assets
Tag what you see.
Boxes, polygons, key points, segmentation — or just a caption with what's in the frame.
Audio612 clips
Capture what you hear.
Read a passage aloud or record what you hear. Transcription, speaker turns, and intent tags follow.
Text2,840 rows
Mark what matters.
Spans, intents, classifications, multi-label. Paste a snippet and tag what matters.
Contribute
As simple as upload and describe.
You don't need a labeling pipeline to contribute. Upload a clip and write what it shows. Read a passage aloud. Tag what's in a photo. The platform turns plain descriptions into structured labels other people can review and reuse.
- 1
Upload anything
A clip, image, recording, or text snippet — under a minute is fine.
- 2
Describe in plain words
What is it, what's said, where is it from. No taxonomy needed.
- 3
It gets structure
Reviewers add labels and tags. You keep credit, license stays yours.
new contribution · video
draft
00:14 / 01:32
A street vendor in Lagos preparing jollof rice in a single iron pot over open flame.
“First we wash the rice three times, then we toast the tomato paste until it turns brick-red…”
food
cooking
Lagos
Public square
Real datasets, built in public.
Browse community datasets. Add a sample, suggest a tag, or fork one into your own project.
01:32
CC-BY 4.0
Street food, around the world
by @wanda.k
“…this is jollof rice from a stall in Lagos. The vendor says it cooks in a single iron pot…”
184 612
Botany
Roadside plants, East Africa
by @kiprop.j
acacia
sisal
frangipani
jacaranda
41 2,840
Linguistics
Hausa proverbs, spoken
by @aminata
Kowa ya yi haƙuri… → Whoever is patient, eats the ripe fruit.
27 318
“Livraison rapide mais l'emballage…” → positive · packaging concern
Sentiment
Customer reviews, French + English
by @lila.m
96 4,120
For teams
Run labeling like product.
Set the rules once. The engine handles routing, scoring, and consensus. You manage the work, not the tooling.
- 01
Define the schema
Set labels, instructions, validation rules, and review thresholds per project.
- 02
Route the work
Assign randomly or by skill. Contributors get the right surface and shortcuts for the format.
- 03
Approve and export
Review queues hold submissions until consensus is met. Clean exports go to the modeling team.
Quality
Data that survives review.
Review queues, contributor history, agreement checks, and export readiness are visible before labels move into a dataset. No silent quality drift.
Approved labels stay separated from drafts and disputes.
Reviewers see the source asset alongside every submission.
Contributor accuracy informs the next assignment.