Vernacore
Language and text data

Language and text data, worked by people who speak it

Transcription, translation review, tagging and speech data — the work where a native speaker is the bottleneck and no model can fill the gap.

The work

What we take on

If your task isn't listed, ask anyway — the answer is either yes, or a straight no with a reason.

  • Transcription and timestamping
  • Translation and translation review
  • Speech data collection and validation
  • Text annotation, tagging and classification
  • Transliteration and romanisation checks
  • Prompt and response review for language models

How it's delivered

Formats
JSON, JSONL, CSV, SRT/VTT, or your own schema
Unit of work
Per file, per minute of audio, or per seat-hour
Quality signal
Sample pass rate against your gold set, reported per batch
Turnaround
Agreed per batch at scoping, then tracked against it daily
Honest scoping

When this is a good fit, and when it isn't

Taking work we'd do badly costs us more than turning it down. This is the list we'd give you on the call anyway.

Good fit

  • You have volume in Indian languages and can't hire for it
  • Your existing vendor's output needs a human review pass
  • You need consistent people rather than a rotating crowd

Not us

  • You need a language nobody on our floor speaks — we will tell you rather than guess
  • You need certified legal or medical translation