Price units

Buy data for AI training: the unit behind each price on our board

Updated

A published price means little until you know its unit: an hour of audio, a whole dataset or a year's access.

Ask about the data I need Tell us what you need, how much and for what. We reply with what the board shows and whether a company licensing its own data could fit.

Units the board records

  • Per dataset purchase: 4 companies, sellers' listings and buyers' payouts together.
  • Per hour of audio or video: 1 companies.
  • By subscription: 1 companies.
  • Unit not stated: 18 companies.

Source: Who sells and who buys AI training data: the board

Among the sellers, Nexdata and Opendatabay price per dataset, Defined.ai priced its 2024 plans by hours of speech and AWS Data Exchange lists subscription and pay-as-you-go products. The per-dataset figures micro1 and Troveo print are what they say companies are paid for their data.

Prices given only on a quote

9 companies on the table give a figure only on request, buyers of company data among them. The sellers of the group are Created by Humans, Grepsr, Shaip, Shutterstock, StageZero and Unidata: ask each for the unit, the licence and the exclusivity before comparing their quotes.

Source: Who sells and who buys AI training data: the board

Tell us what training data you need

Tick the box above to send your enquiry.

Sources

This site's own dataset, every figure quoted from its source and dated:

  1. Datarade: Nexdata multilingual read speech 65,000 hours
  2. Datarade: Nexdata call-centre telephony speech 20,000 hours
  3. Datarade: Nexdata speech synthesis 400 hours
  4. Datarade: Nexdata code-switching speech 5,000 hours
  5. Datarade: Nexdata in-car speech 15,000 hours
  6. Datarade: StageZero Lithuanian speech 20 hours
  7. Datarade: Grepsr tailored web data
  8. Datarade: audio datasets category

Looking for a kind of data the board does not list?

Tell us what you need, how much and for what. We reply with what the board shows and whether a company licensing its own data could fit.

Ask about the data I need