Speakers
5
The data we capture.
Six types of human signal models can't generate for themselves. Captured from real people doing real work. Produced to your brief, not pulled off a shelf.
How data capture works.
Every dataset is designed, screened, and quality-checked before it ships, not pulled from a queue.
01
Brief
You bring the capability gap. We turn it into a data spec.
02
Design & screen
We design the task and screen the right people, credentials checked where it matters.
03
Capture
Facilitated, instrumented sessions. Consent and provenance recorded at the source.
04
QC & adjudicate
Sessions are reviewed; weak or off-brief ones are rejected and re-run.
05
Deliver
Schema-conformant to your pipeline, with the full provenance chain.
What a deliverable looks like.
Structured, diarised, and provenance-stamped. Schema-conformant to your spec, ready to ingest.
A conversational voice dataset: five speakers across three accent profiles, paired negotiations and solo speech, word-aligned and diarised. Every clip carries interaction metrics and a signed consent chain.
Accent profiles
3
Speaker changes
79
Median gap
833ms
Overlap
5.0%
Manifest
NeMo / HF
Per-speaker Opus/OGG · word-aligned manifest · Annature consent envelopes · IP assigned at capture · checksummed. Identities redacted.
RECRUITMENT & DELIVERY INFRASTRUCTURE
Built on eight years of Askable infrastructure.
Recruitment, consent, capture, transcription, and delivery run inside one audited platform: one identity system, one consent ledger, one audit log. Provenance and consent are captured at the source, not reconstructed after.
Research sessions run
230,121& counting
Participant panel
50+ countries5,300+ cities · triple-verified
Show rate
97.8%63% above industry average
Operating since
2017years of production engineering
Compliance
ISO + SOC 227001 · 27701 · 42001 · Type II
One platform, one audit trail. Granular consent, attribution chains, and audit logs on every artifact. Independent: not owned by, or exclusive to, any lab.
See the audit postureSTART WITH A GOAL
Tell us the capability you're trying to move.
Bring the model behaviour you want to improve. We'll scope the data, run a validated pilot, and deliver it ingestion-ready.