Speech and language data prices, seller by seller
12 sellers of speech and language data were read for the prices they publish and what they pay contributors, each answer quoted from their own pages, listings and announcements, read 22 October 2026.
How the data is priced
The published price, in its words
What kind of speech or text
What the contributor is paid
All the answers, seller by seller
Every row is a single seller, and every answer expands to the wording it was taken from, the page it came from and when it was read. "Not stated" is used when the page says nothing on the point.
| Seller | How the data is priced | The published price, in its words | What kind of speech or text | What the contributor is paid |
|---|---|---|---|---|
| Appen | not statedNot addressed in Appen off-the-shelf datasets catalogue | not statedNot addressed in Appen off-the-shelf datasets catalogue | not statedNot addressed in Appen off-the-shelf datasets catalogue | not statedNot addressed in Appen off-the-shelf datasets catalogue |
| Bartz v. Anthropic settlement | a price per book or per work | It provides approximately $3,000 per work (not per Class Member), plus interest earned on… | books | The Settlement provides that for many types of works such as trade books and university… |
| Created by Humans | not statedNot addressed in Created by Humans home page | not statedNot addressed in Created by Humans home page | not statedNot addressed in Created by Humans home page | Get compensated when your work is licensed |
| Datoric AI | not statedNot addressed in tts-conversational-voice-20000h dataset card | not statedNot addressed in tts-conversational-voice-20000h dataset card | conversational speech | not statedNot addressed in tts-conversational-voice-20000h dataset card |
| Defined.ai | a subscription or access planNot addressed in Defined.ai audio datasets page | Standard Plan: 315 hours of speech data for €50,000 (30% discount)Not addressed in Defined.ai audio datasets page | several kinds Not addressed in Data Access Plan announcement | not statedNot addressed in Defined.ai audio datasets page Not addressed in Data Access Plan announcement |
| HarperCollins | a price per book or per work | The HarperCollins licensing deal provides for a $5,000 fee per title, which will be split… | books | which will be split 50-50 between an author who chooses to participate and HarperCollins |
| Linguistic Data Consortium | a price per dataset or corpus | Non-members may license this data for $3000. | not statedNot addressed in LDC new corpora announcement (LINGUIST List) | not statedNot addressed in LDC new corpora announcement (LINGUIST List) |
| Nexdata | a price per dataset or corpus | License Starts at One-off purchase $20,000 / purchase | varies by document Not addressed in Speech synthesis data collection services (Datarade listing) | not statedNot addressed in Multilingual read speech data, 65,000 hours (Datarade listing) Not addressed in Unscripted call centre telephony speech, 20,000 hours (Datarade listing) Not addressed in Speech synthesis data collection services (Datarade listing) Not addressed in 8kHz conversational speech, 15,000 hours (Datarade listing) Not addressed in Code-switching speech, 5,000 hours (Datarade listing) |
| Opendatabay | not statedNot addressed in Opendatabay marketplace home | not statedNot addressed in Opendatabay marketplace home | not statedNot addressed in Opendatabay marketplace home | not statedNot addressed in Opendatabay marketplace home |
| Princeton University Press | not statedNot addressed in Princeton University Press statement on the Bartz v. Anthropic settlement | not statedNot addressed in Princeton University Press statement on the Bartz v. Anthropic settlement | books | Princeton University Press is fully aligned with the Settlement’s default split of 50/50… |
| Shaip | priced on request Not addressed in Shaip speech data collection page | not statedNot addressed in Shaip TTS datasets page Not addressed in Shaip speech data collection page | several kindsNot addressed in Shaip TTS datasets page | not statedNot addressed in Shaip TTS datasets page Not addressed in Shaip speech data collection page |
| Wirestock | not statedNot addressed in Wirestock home page | not statedNot addressed in Wirestock home page | not statedNot addressed in Wirestock home page | Upload your work once and earn commission whenever it is licensed. |
Each answer in its source's own words (72 quotes)
- Appen, Appen off-the-shelf datasets catalogue, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 6,844 characters long.
- Appen, Appen off-the-shelf datasets catalogue, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 6,844 characters long.
- Appen, Appen off-the-shelf datasets catalogue, read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 6,844 characters long.
- Appen, Appen off-the-shelf datasets catalogue, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 6,844 characters long.
- Bartz v. Anthropic settlement, Settlement website, frequently asked questions, read 2026-10-22, How the data is priced:
It provides approximately $3,000 per work (not per Class Member)
- Bartz v. Anthropic settlement, Settlement website, frequently asked questions, read 2026-10-22, The published price, in its words:
It provides approximately $3,000 per work (not per Class Member), plus interest earned on the Settlement fund, less the Court-approved costs and fees taken out of the Settlement fund.
- Bartz v. Anthropic settlement, Settlement website, frequently asked questions, read 2026-10-22, What kind of speech or text:
This means in most cases that you are a publisher or author of a book
- Bartz v. Anthropic settlement, Settlement website, frequently asked questions, read 2026-10-22, What the contributor is paid:
The Settlement provides that for many types of works such as trade books and university press books, the payment, by default, will be split 50-50 between authors and publishers.
- Created by Humans, Created by Humans home page, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 3,758 characters long.
- Created by Humans, Created by Humans home page, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 3,758 characters long.
- Created by Humans, Created by Humans home page, read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 3,758 characters long.
- Created by Humans, Created by Humans home page, read 2026-10-08, What the contributor is paid:
Get compensated when your work is licensed
- Datoric AI, tts-conversational-voice-20000h dataset card, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 6,021 characters long.
- Datoric AI, tts-conversational-voice-20000h dataset card, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 6,021 characters long.
- Datoric AI, tts-conversational-voice-20000h dataset card, read 2026-10-08, What kind of speech or text:
20,000 hours of 48kHz conversational speech across 30+ languages
- Datoric AI, tts-conversational-voice-20000h dataset card, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 6,021 characters long.
- Defined.ai, Data Access Plan announcement, read 2026-10-08, How the data is priced:
Defined.ai Launches Data Access Plan (DAP): Allowing Access to Speech Datasets at affordable prices
- Defined.ai, Data Access Plan announcement, read 2026-10-08, The published price, in its words:
Standard Plan: 315 hours of speech data for €50,000 (30% discount)
- Defined.ai, Data Access Plan announcement, read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 5,070 characters long.
- Defined.ai, Data Access Plan announcement, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 5,070 characters long.
- Defined.ai, Defined.ai audio datasets page, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 4,578 characters long.
- Defined.ai, Defined.ai audio datasets page, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 4,578 characters long.
- Defined.ai, Defined.ai audio datasets page, read 2026-10-08, What kind of speech or text:
from on-device scripted speech to call-center-style dialogue, live podcasts, music tracks and licensed SFX
- Defined.ai, Defined.ai audio datasets page, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 4,578 characters long.
- HarperCollins, Authors Guild on the HarperCollins AI licensing deal, read 2026-10-22, How the data is priced:
The HarperCollins licensing deal provides for a $5,000 fee per title
- HarperCollins, Authors Guild on the HarperCollins AI licensing deal, read 2026-10-22, The published price, in its words:
The HarperCollins licensing deal provides for a $5,000 fee per title, which will be split 50-50 between an author who chooses to participate and HarperCollins.
- HarperCollins, Authors Guild on the HarperCollins AI licensing deal, read 2026-10-22, What kind of speech or text:
publishers must obtain permission from authors before including their books in AI licensing deals
- HarperCollins, Authors Guild on the HarperCollins AI licensing deal, read 2026-10-22, What the contributor is paid:
which will be split 50-50 between an author who chooses to participate and HarperCollins
- Linguistic Data Consortium, LDC new corpora announcement (LINGUIST List), read 2026-10-08, How the data is priced:
2026 members can access this corpus through their LDC accounts. Non-members may license this data for $3000.
- Linguistic Data Consortium, LDC new corpora announcement (LINGUIST List), read 2026-10-08, The published price, in its words:
Non-members may license this data for $3000.
- Linguistic Data Consortium, LDC new corpora announcement (LINGUIST List), read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 6,265 characters long.
- Linguistic Data Consortium, LDC new corpora announcement (LINGUIST List), read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 6,265 characters long.
- Nexdata, Code-switching speech, 5,000 hours (Datarade listing), read 2026-10-08, How the data is priced:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Code-switching speech, 5,000 hours (Datarade listing), read 2026-10-08, The published price, in its words:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Code-switching speech, 5,000 hours (Datarade listing), read 2026-10-08, What kind of speech or text:
AI Training DataAudio DataMixed Speech Data |5,000 Hours |Code-switching|Audio Data| Speech R...
- Nexdata, Code-switching speech, 5,000 hours (Datarade listing), read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 9,823 characters long.
- Nexdata, 8kHz conversational speech, 15,000 hours (Datarade listing), read 2026-10-08, How the data is priced:
License Starts at One-off purchase $100,000 / purchase
- Nexdata, 8kHz conversational speech, 15,000 hours (Datarade listing), read 2026-10-08, The published price, in its words:
License Starts at One-off purchase $100,000 / purchase
- Nexdata, 8kHz conversational speech, 15,000 hours (Datarade listing), read 2026-10-08, What kind of speech or text:
AI Training DataAudio DataConversation Audio Data8kHz Conversational Speech Data | 15,000 Hours | Audio Data | Speec...
- Nexdata, 8kHz conversational speech, 15,000 hours (Datarade listing), read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 10,139 characters long.
- Nexdata, Multilingual read speech data, 65,000 hours (Datarade listing), read 2026-10-08, How the data is priced:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Multilingual read speech data, 65,000 hours (Datarade listing), read 2026-10-08, The published price, in its words:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Multilingual read speech data, 65,000 hours (Datarade listing), read 2026-10-08, What kind of speech or text:
Multilingual Read Speech Data
- Nexdata, Multilingual read speech data, 65,000 hours (Datarade listing), read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 10,184 characters long.
- Nexdata, Speech synthesis data collection services (Datarade listing), read 2026-10-08, How the data is priced:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Speech synthesis data collection services (Datarade listing), read 2026-10-08, The published price, in its words:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Speech synthesis data collection services (Datarade listing), read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 10,778 characters long.
- Nexdata, Speech synthesis data collection services (Datarade listing), read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 10,778 characters long.
- Nexdata, Unscripted call centre telephony speech, 20,000 hours (Datarade listing), read 2026-10-08, How the data is priced:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Unscripted call centre telephony speech, 20,000 hours (Datarade listing), read 2026-10-08, The published price, in its words:
License Starts at One-off purchase $20,000 / purchase
- Nexdata, Unscripted call centre telephony speech, 20,000 hours (Datarade listing), read 2026-10-08, What kind of speech or text:
AI Training DataAudio DataCall Center Audio DataUnscripted Call Center Telephony Speech Data | 20,000 Hours |Speech...
- Nexdata, Unscripted call centre telephony speech, 20,000 hours (Datarade listing), read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 10,525 characters long.
- Opendatabay, Opendatabay marketplace home, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 8,326 characters long.
- Opendatabay, Opendatabay marketplace home, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 8,326 characters long.
- Opendatabay, Opendatabay marketplace home, read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 8,326 characters long.
- Opendatabay, Opendatabay marketplace home, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 8,326 characters long.
- Princeton University Press, Princeton University Press statement on the Bartz v. Anthropic settlement, read 2026-10-22, How the data is priced: The page is silent here; it was read in full on 2026-10-22, 14,330 characters long.
- Princeton University Press, Princeton University Press statement on the Bartz v. Anthropic settlement, read 2026-10-22, The published price, in its words: The page is silent here; it was read in full on 2026-10-22, 14,330 characters long.
- Princeton University Press, Princeton University Press statement on the Bartz v. Anthropic settlement, read 2026-10-22, What kind of speech or text:
the thousands of PUP and Island Press titles used without permission for training LLMs are part of the settlement
- Princeton University Press, Princeton University Press statement on the Bartz v. Anthropic settlement, read 2026-10-22, What the contributor is paid:
Princeton University Press is fully aligned with the Settlement’s default split of 50/50 between authors and publishers, for trade and university presses.
- Shaip, Shaip speech data collection page, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 11,966 characters long.
- Shaip, Shaip speech data collection page, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 11,966 characters long.
- Shaip, Shaip speech data collection page, read 2026-10-08, What kind of speech or text:
Build a text-to-speech (TTS) multilingual model with the help of our global workforce, who help you collect speech data in 150+ languages
- Shaip, Shaip speech data collection page, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 11,966 characters long.
- Shaip, Shaip TTS datasets page, read 2026-10-08, How the data is priced:
Licensing: Get a license tailored to your needs.
- Shaip, Shaip TTS datasets page, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 3,285 characters long.
- Shaip, Shaip TTS datasets page, read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 3,285 characters long.
- Shaip, Shaip TTS datasets page, read 2026-10-08, What the contributor is paid: The page is silent here; it was read in full on 2026-10-08, 3,285 characters long.
- Wirestock, Wirestock home page, read 2026-10-08, How the data is priced: The page is silent here; it was read in full on 2026-10-08, 4,468 characters long.
- Wirestock, Wirestock home page, read 2026-10-08, The published price, in its words: The page is silent here; it was read in full on 2026-10-08, 4,468 characters long.
- Wirestock, Wirestock home page, read 2026-10-08, What kind of speech or text: The page is silent here; it was read in full on 2026-10-08, 4,468 characters long.
- Wirestock, Wirestock home page, read 2026-10-08, What the contributor is paid:
Upload your work once and earn commission whenever it is licensed.
Documents we read but did not count
9 named documents were read but are not the kind of document this board answers from; see why below. They are counted and never answered.
- Linguistic Data Consortium - Linguistic Data Consortium: the catalogue page stored only its title; no prices or terms
- ELRA Catalogue of Language Resources: a catalogue entry list; no price or terms are in the text read
- Language Resources Announcements | ELRA: a list of new-resource announcements; no price or terms are in the text read
- https://www.bas.uni-muenchen.de/forschung/Bas/BasPricingdeu.pdf: the read stored no text (an empty PDF read)
- 100,000th Job Posted to Voices in its 10th Year of Business - Voices.com | Voices: a Voices.com press item about a job-posting milestone, not a price or contributor term for data
- datarade.ai: a vendor profile with no price or contributor terms in the text read
- Common Voice on Hugging Face (CC0 licence card): the card says Common Voice moved to Mozilla Data Collective; no price or licence terms remain on it
- Lanfrica, the shared evidence layer for African AI: a directory and news page for African AI; it sells no data and states no price
- Wiley: AI demand drives first quarter fiscal 2026 results: a quarterly results release: AI licensing revenue in aggregate, with no price per title, no unit and no author terms
The method in full
Each seller's own catalogue, pricing, contributor and book licensing pages were read through estate-fetch and stored whole (sha256 on every row).
For every page, four questions were answered from its own words: how the data is priced, the published price, in its words, what kind of speech or text, what the contributor is paid.
Every stated answer carries the sentence it rests on, checked against the stored text; 'not stated' means the page read does not address the question.
A seller with several pages reads 'varies' where they disagree.
12 sellers, 18 pages.
5 named documents wait to be read; they are left out of every count.
Where we looked (32)
- Audio datasets for AI | ASR | TTS | Music | Defined.ai (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Linguistic Data Consortium - Linguistic Data Consortium (read, set aside)
- Data Access Plan (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- DatoricAI/tts-conversational-voice-20000h · Datasets at Hugging Face (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- LINGUIST List: LDC new corpora announcement (non-member fees) (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- ELRA Catalogue of Language Resources (read, set aside)
- Language Resources Announcements | ELRA (read, set aside)
- https://www.bas.uni-muenchen.de/forschung/Bas/BasPricingdeu.pdf (read, set aside)
- datarade.ai (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- 100,000th Job Posted to Voices in its 10th Year of Business - Voices.com | Voices (read, set aside)
- https://tts.ai/sell-voice/?lang=en (not yet read)
- Text To Speech Datasets | TTS Data for ASR and NLP Projects (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- datarade.ai (read, set aside)
- datarade.ai (not yet read)
- Speech Data Collection Services | Audio & Voice Training Data For AI (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- AI training datasets: off-the-shelf data catalog | Appen (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Earn Money Creating AI Training Data | Wirestock (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Licensed AI Training Data & LLM Fine-Tuning Datasets | Opendatabay (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Common Voice on Hugging Face (CC0 licence card) (read, set aside)
- Lanfrica, the shared evidence layer for African AI (read, set aside)
- The AI Rights licensing platform for creators - Created by Humans (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- datarade.ai (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- datarade.ai (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- datarade.ai (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- datarade.ai (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- the universe, rule 1 (not yet read)
- Linguistic Data Consortium, Join LDC (own membership fees) (not yet read)
- Linguistic Data Consortium, Obtaining Data (own licensing terms) (not yet read)
- Bartz v. Anthropic settlement: frequently asked questions (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Authors Guild: the HarperCollins AI licensing deal (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Princeton University Press statement on the Bartz v. Anthropic settlement (cited for reporting; the publisher's own catalogue, pricing, contributor and book licensing pages)
- Wiley: AI demand drives first quarter fiscal 2026 results (read, set aside)
Take this figure with you
12 sellers of speech and language data were read for the prices they publish and what they pay contributors, each answer quoted from their own pages, listings and announcements, read 22 October 2026.
How the data is priced, by seller: not stated 6 of 12, a price per dataset or corpus 2 of 12, a price per book or per work 2 of 12, priced on request 1 of 12, a subscription or access plan 1 of 12 (12 sellers, read 22 October 2026).
The published price, in its words, by seller: not stated 7 of 12, stated 5 of 12 (12 sellers, read 22 October 2026).
What kind of speech or text, by seller: not stated 5 of 12, books 3 of 12, several kinds 2 of 12, conversational speech 1 of 12, varies 1 of 12 (12 sellers, read 22 October 2026).
What the contributor is paid, by seller: not stated 7 of 12, stated 5 of 12 (12 sellers, read 22 October 2026).
Citation: "Speech and language data prices, seller by seller", updated 2026-10-22, https://languagedatalicensing.com/datasets/language-data-rates-by-language/.
Save the CSV (72 answers; derived 2026-10-22)