Audio and speech · Speech in languages other than English · Updated 30 Oct 2026

Who buys speech recordings in languages other than English, and what do they pay per hour?

AIData.Broker represents the companies and people that hold speech in other languages when they sell to AI labs: we value what they hold against public prices, describe it without their name to every buyer that takes it at the same time and help negotiate the terms. Our fee is 12% of the first $250,000 a sale pays, and nothing if no sale closes.

Calls, conversations, broadcasts and read speech in Spanish, French, Arabic, Hindi and the rest. Speech and voice models are weakest outside English, so buyers price recordings by language and by how few exist.

Value your speech in other languagesSelling audio and speech recordings for AI training

No data upload
About two minutes
No email needed

Prices on record: 137
sales that happened, offered to holders, asking prices off the shelf, made to order, cost to produce
Who holds it: Call centres serving Spanish, French or other markets, Radio and TV in other languages, Transcription and translation companies
Who has paid, offered or listed: Karya, Beijing Haitian Ruisheng Science Technology (SpeechOcean), Shaip, Opendatabay, Infobay.AI

The short answer · figures from the market's records, last reviewed 30 Oct 2026

Who buys speech recordings in languages other than English, and what do they pay per hour?

Speech and voice-agent companies, and the catalogues that resell to them, priced by language and by hour. Mercor's listing for call centres asks for recorded calls ideally in more than one language and names U.S. Hispanic and francophone markets; research catalogues license telephone speech for commercial use, such as Spanish SpeechDat at €19,500 for 247 hours from ELRA; and Shaip offers up to Rs. 200 per audio hour for audio a holder already has in Indian languages. What platforms pay speakers to record new audio sets what an hour costs a buyer to have made: Mercor paid Arabic voice actors $175 per finished hour.

What buyers have paid, offered or asked for speech in other languages

The whole price board

Every price in the market's records about speech in other languages, each with its source and the day it was read. Open a row for the source's own words.

137 public prices

One hour of Odia (low-resource Indian language) speech data
Karya co-founder Safiya Husain's figure for what buyers now pay; a tenfold rise as big tech sought Indian-language data.
Audio and speechSales that happenedDec 2023Press report
Context (Thomson Reuters Foundation)
About $3.00 to $40
per hour of audio
$40 per hour of audio
Off-the-shelf speech recognition training datasets (wake words, read sentences, dialogues; Chinese and foreign languages), usage licences sold by Haitian Ruisheng in 2020: 198,951 hours
Realised sales: the seller licenses the use of datasets it owns (one dataset can be licensed many times), and the prospectus reports hours sold and the average price per hour for 2020, audited for the IPO. These are pre-generative-AI speech corpora sold mostly to Chinese and foreign tech firms; read them as the price history of off-the-shelf speech licensing, not a 2026 rate. The same filing splits 2020 by language: Chinese 91,455 hours at CNY 255 an hour, foreign languages 108,525 hours at CNY 674 an hour (all speech, both kinds).
Audio and speechSales that happenedDec 2020The payer's own page
Beijing Haitian Ruisheng Science Technology (SpeechOcean), STAR Market IPO prospectus (招股说明书), 9 Aug 2021, training data products table
$60
per hour, average across all speech recognition dataset hours licensed in 2020
Published as 413 元/小时 (CNY 413 an hour, average over 198,951 hours sold)
Off-the-shelf speech recognition training datasets (wake words, read sentences, dialogues; Chinese and foreign languages), usage licences sold by Haitian Ruisheng in 2019: 136,411 hours
Realised sales: the seller licenses the use of datasets it owns (one dataset can be licensed many times), and the prospectus reports hours sold and the average price per hour for 2019, audited for the IPO. These are pre-generative-AI speech corpora sold mostly to Chinese and foreign tech firms; read them as the price history of off-the-shelf speech licensing, not a 2026 rate. The same filing splits 2019 by language: Chinese 55,957 hours at CNY 317 an hour, foreign languages 80,728 hours at CNY 793 an hour.
Audio and speechSales that happenedDec 2019The payer's own page
Beijing Haitian Ruisheng Science Technology (SpeechOcean), STAR Market IPO prospectus (招股说明书), 9 Aug 2021, training data products table
$79
per hour, average across all speech recognition dataset hours licensed in 2019
Published as 549 元/小时 (CNY 549 an hour, average over 136,411 hours sold)
Off-the-shelf speech recognition training datasets (wake words, read sentences, dialogues; Chinese and foreign languages), usage licences sold by Haitian Ruisheng in 2018: 115,636 hours
Realised sales: the seller licenses the use of datasets it owns (one dataset can be licensed many times), and the prospectus reports hours sold and the average price per hour for 2018, audited for the IPO. These are pre-generative-AI speech corpora sold mostly to Chinese and foreign tech firms; read them as the price history of off-the-shelf speech licensing, not a 2026 rate.
Audio and speechSales that happenedDec 2018The payer's own page
Beijing Haitian Ruisheng Science Technology (SpeechOcean), STAR Market IPO prospectus (招股说明书), 9 Aug 2021, training data products table
$76
per hour, average across all speech recognition dataset hours licensed in 2018
Published as 504 元/小时 (CNY 504 an hour, average over 115,636 hours sold)
Existing Indic-language audio a holder already has (podcasts, radio programmes, news broadcasts, talk shows; published or archived), licensed to Shaip for AI training only
The only buyer page read that states a per-hour price for audio a holder already has; the licence is for training only, the content is not published or redistributed and ownership stays with the holder. Media houses and radio channels are named among those it partners with, and the page pays 5% of the contract value to whoever brings in a holder. Undated page (footer 2026).
Audio and speechOffered to holdersOct 2026The payer's own page
Shaip Crowd, Indic Language Voice Data Collection Project
Up to $2.11
per hour of audio
Hindi call-center recordings, dual-channel, 1,587,658 hours (Infobay.AI on Opendatabay)
Infobay.AI's asking price on the Opendatabay marketplace (prices in USD since 1 October 2026, converted from GBP at $1 = 0.76 GBP); the listing says final pricing depends on hours, metadata and annotation, so it is a list price, not a quote. Repriced on 24 Oct 2026 (the listing says 'UPDATED 24/10/2026').
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Opendatabay marketplace, 1.5Mn+ Hours Hindi Dual-Channel Call Center Dataset
$526K
per dataset of 1,587,658 hours
$0.33 per hour of audio
Hindi podcast audio, dual-channel, 11,607 hours (Infobay.AI on Opendatabay)
Infobay.AI's asking price on the Opendatabay marketplace (prices in USD since 1 October 2026, converted from GBP at $1 = 0.76 GBP); the listing says final pricing depends on hours, metadata and annotation, so it is a list price, not a quote. Repriced on 24 Oct 2026 (the listing says 'UPDATED 24/10/2026').
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Opendatabay marketplace, 11K+ Hours Hindi Dual-Channel Podcast Audio Dataset
$64.2K
per dataset of 11,607 hours
$5.53 per hour of audio
Swahili call-center recordings, dual-channel, 194,331 hours (Infobay.AI on Opendatabay)
Infobay.AI's asking price on the Opendatabay marketplace (prices in USD since 1 October 2026, converted from GBP at $1 = 0.76 GBP); the listing says final pricing depends on hours, metadata and annotation, so it is a list price, not a quote. Repriced on 24 Oct 2026 (the listing says 'UPDATED 24/10/2026').
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Opendatabay marketplace, 194K+ Hours Swahili Dual-Channel Call Center Dataset
$316K
per dataset of 194,331 hours
$1.63 per hour of audio
GlobalPhone Arabic (Karlsruhe Institute of Technology): about 20 hours of read Modern Standard Arabic political news, about 100 sentences from each of about 100 speakers (ELRA catalogue ELRA-S0192)
ELRA's non-member price for a company under the commercial-use (VAR) licence; an academic non-commercial licence is EUR 700, and buying several GlobalPhone languages earns special prices. Read news sentences, so it prices clean read MSA, not dialect.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, GlobalPhone Arabic (ELRA-S0192)
$4,029
per corpus of 20 hours
Published as 3600.00 € (non-member, commercial use)
$201 per hour of audio
NEMLAR Broadcast News Speech Corpus: 40 hours of Arabic radio broadcast news (ELRA catalogue ELRA-S0219)
ELRA's non-member price for a company under the commercial-use (VAR) licence; an academic non-commercial licence is EUR 300. Discounts of 15% and 30% when bought with the other NEMLAR resources. Broadcast news, so mostly MSA.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, NEMLAR Broadcast News Speech Corpus (ELRA-S0219)
$4,477
per corpus of 40 hours
Published as 4000.00 € (non-member, commercial use)
$112 per hour of audio
NetDC Arabic Broadcast News Speech Corpus: about 22.5 hours of Standard Arabic radio news from Radio Orient (France), November 2001 to January 2002, about 90 speakers (ELRA catalogue ELRA-S0157)
ELRA's non-member price for a company under the commercial-use (VAR) licence; an academic non-commercial licence is EUR 200. Twin of LDC's VOA Arabic set ($700 for 10 hours, research only).
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, NetDC Arabic BNSC (Broadcast News Speech Corpus) (ELRA-S0157)
$3,022
per corpus of 22.5 hours
Published as 2700.00 € (non-member, commercial use)
$134 per hour of audio
OrienTel Egypt MSA: Modern Standard Arabic read by Egyptians over fixed and mobile telephone networks, 50 hours from 500 speakers (ELRA catalogue ELRA-S0222)
ELRA's non-member price for a company under the commercial-use (VAR) licence; prompted telephone items in MSA. The Jordan MSA set (ELRA-S0290: 556 speakers, 50 hours) is priced the same at EUR 20,000, so it is not a separate row; the colloquial Egypt and Jordan sets cost about three times as much.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, OrienTel Egypt MSA (Modern Standard Arabic) database (ELRA-S0222)
$22.4K
per corpus of 50 hours
Published as 20000.00 € (non-member, commercial use)
$448 per hour of audio
OrienTel Morocco MCA: colloquial Moroccan Arabic (Darija) over fixed and mobile telephone networks, 53 hours from 772 speakers, prompted items (ELRA catalogue ELRA-S0183)
ELRA's non-member price for a company under the commercial-use (VAR) licence. A quarter of what ELRA asks for the Jordan colloquial set (EUR 60,250 for 50 hours); the Morocco MCA, MSA and French-as-spoken-in-Morocco sets are sold as a bundle for EUR 27,500 (non-member, commercial). Prompted telephone items (digits, dates, names), not free conversation.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, OrienTel Morocco MCA (Modern Colloquial Arabic) database (ELRA-S0183)
$16.8K
per corpus of 53 hours
Published as 15000.00 € (non-member, commercial use)
$317 per hour of audio
OrienTel Morocco MSA: Modern Standard Arabic read by Moroccans over fixed and mobile telephone networks, 49 hours from 530 speakers (ELRA catalogue ELRA-S0184)
ELRA's non-member price for a company under the commercial-use (VAR) licence; prompted telephone items in MSA. The Moroccan colloquial set costs EUR 15,000 for 53 hours, so dialect sells at about 1.4 times MSA from the same project. The Tunisia MSA set (ELRA-S0187: 598 speakers, 50 hours) is the same price, EUR 10,000, so it is not a separate row.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, OrienTel Morocco MSA (Modern Standard Arabic) database (ELRA-S0184)
$11.2K
per corpus of 49 hours
Published as 10000.00 € (non-member, commercial use)
$228 per hour of audio
UAE Arabic Speech Recognition Corpus (Mobile): 85.65 hours of read news and daily-dialogue scripts from 168 speakers (94 men, 74 women) in the UAE, 2 channels, 16 kHz, quiet office or home (ELRA catalogue ELRA-S0228_130)
ELRA's price is the same EUR 36,000 for members and non-members and for research or commercial use; available since 28 June 2024. Read speech from screened speakers with clear pronunciation, so it prices Emirati voices rather than Emirati conversation; the only priced UAE speech corpus found.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, UAE Arabic Speech Recognition Corpus (Mobile) (ELRA-S0228_130)
$40.3K
per corpus of 85.65 hours
Published as 36000.00 € (non-member, commercial use)
$470 per hour of audio
BBN/AUB DARPA Babylon Levantine Arabic Speech and Transcripts: 60.6 hours of spontaneous responses to prompts from 164 speakers (101 men, 63 women) recorded in Boston and Beirut in 2002
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $1,250.00. Microphone speech from paid subjects in a refugee and medical speech-translation scenario, with transcripts included; the dearest Levantine set per hour in LDC's catalogue.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, BBN/AUB DARPA Babylon Levantine Arabic Speech and Transcripts (LDC2005S08)
$2,500
per corpus of 60.6 hours
$41 per hour of audio
CALLHOME Egyptian Arabic Speech: 120 unscripted telephone conversations of up to 30 minutes between native speakers of Egyptian Colloquial Arabic (Cairene)
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $750.00. Calls lasted up to 30 minutes, but the page states no total, so no per-hour figure. Speech only: transcripts (LDC97T19) and a lexicon (LDC99L22) are sold separately.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, CALLHOME Egyptian Arabic Speech (LDC97S45)
$1,500
per corpus of 120 conversations (total hours not stated)
Gulf Arabic Conversational Telephone Speech (collected by Appen in 2004): about 2,800 minutes of spontaneous colloquial Gulf Arabic telephone conversations from 975 speakers, 8 kHz
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $400.00. The only Gulf Arabic speech in LDC's catalogue; transcripts are sold separately (LDC2006T15). Telephone audio recorded in 2004, so it says little about what fresh Gulf conversation sells for.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, Gulf Arabic Conversational Telephone Speech (LDC2006S43)
$800
per corpus of about 2,800 minutes (46.7 hours)
$17 per hour of audio
Iraqi Arabic Conversational Telephone Speech (collected by Appen in 2003 and 2004): about 3,000 minutes of spontaneous colloquial Iraqi Arabic telephone conversations from 474 speakers, 8 kHz
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $200.00. Transcripts are sold separately (LDC2006T16). The cheapest Arabic dialect speech per hour in LDC's catalogue.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, Iraqi Arabic Conversational Telephone Speech (LDC2006S45)
$400
per corpus of about 3,000 minutes (50 hours)
$8.00 per hour of audio
IWSLT 2022-2023 Shared Task set: 210 hours of Tunisian Arabic conversational telephone speech (1,188 two-channel conversations recorded in Tunis 2016-2017 with informed consent), transcripts and English translations of 175 hours
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $1,250.00. Released 2025, so it is the freshest dialect conversation LDC sells; the price includes transcripts and English translations, which makes the per-hour figure a floor for transcribed dialect speech at research rates.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, IWSLT 2022-2023 Shared Task Training, Development and Test Set (LDC2025S05)
$2,500
per corpus of 210 hours
$12 per hour of audio
Levantine Arabic Conversational Telephone Speech (collected and transcribed by Appen in 2004): 985 conversation sides from 982 speakers, 5 to 6 minutes a side
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $200.00. The page gives sides and the average length of a side but not total hours, so no per-hour figure is derived (two sides can share one recording). Sister corpus to the Gulf and Iraqi Appen sets at $800 and $400.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, Levantine Arabic Conversational Telephone Speech (LDC2007S01)
$400
per corpus (985 conversation sides; total hours not stated)
Levantine Arabic QT Training Data Set 5, Speech: about 250 hours of Levantine Arabic telephone conversation (1,660 calls, 2003 to 2005, mostly Lebanese speakers)
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $2,000.00. Set 5 combines LDC's earlier Levantine sets, which LDC still sells on their own: Fisher Training Data Set 3 (LDC2005S07, about 50 hours, $1,000, so $20 an hour) and QT Set 4 (LDC2005S14, about 138 hours with transcripts, $3,000, so $21.74 an hour); one row here because the audio overlaps. Transcripts for Set 5 are sold separately (LDC2006T07).
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, Levantine Arabic QT Training Data Set 5, Speech (LDC2006S29)
$4,000
per corpus of about 250 hours
$16 per hour of audio
Multi-Language Conversational Telephone Speech 2011, Arabic Group: about 117 hours of telephone calls in colloquial Iraqi, Levantine and Maghrebi Arabic
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $1,250.00. Callers phoned acquaintances in their own social network for up to 15 minutes each; no transcripts, and the page does not split the hours by dialect. Collected for language identification (NIST LRE 2011).
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, Multi-Language Conversational Telephone Speech 2011 -- Arabic Group (LDC2019S02)
$2,500
per corpus of about 117 hours
$21 per hour of audio
2019 NIST Speaker Recognition Evaluation Test Set, CTS Challenge: about 635 hours of Tunisian Arabic telephone recordings (Call My Net 2 collection) with answer keys
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $200.00. An evaluation set for speaker recognition, not a transcribed training corpus, which is why it sells under a dollar an hour; the same Tunisian calls feed the 2018 SRE test set (LDC2020S04, $750 for 396 hours mixed with English video).
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, 2019 NIST Speaker Recognition Evaluation Test Set -- CTS Challenge (LDC2023S03)
$400
per corpus of about 635 hours
$0.63 per hour of audio
Arabic Broadcast News Speech: 10 hours of Voice of America Arabic satellite radio news recorded 2000 to 2001, 16 kHz (transcripts sold separately)
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $350.00. Mostly Modern Standard Arabic news reading.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, Arabic Broadcast News Speech (LDC2006S46)
$700
per corpus of 10 hours
$70 per hour of audio
West Point Arabic Speech: 11.42 hours of read Modern Standard Arabic prompts from 110 speakers (75 native), 22.05 kHz microphone recordings
LDC's one-time non-member licence fee, shown under View Fees; LDC licences to non-members exclude commercial use, which needs a for-profit membership. The reduced-licence fee is $500.00. Read prompts recorded by the US Military Academy for language teaching; small and read, which is why it costs several times the telephone conversation sets per hour.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, West Point Arabic Speech (LDC2002S02)
$1,000
per corpus of 11.42 hours
$88 per hour of audio
Arabic (Darija) conversational speech dataset recorded in Morocco: 1,500 hours of spontaneous two-person conversations from 150 native speakers, dual-channel, time-aligned human-verified transcripts, signed commercial-use consent
Off-the-shelf list price for a non-exclusive commercial training licence; volume discounts on request. The full 1,500 hours lists at $97,500. ELRA's prompted Moroccan telephone set costs about five times as much per hour.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
SpeechData.ai catalogue, Arabic (Darija) conversational speech dataset
$65
per hour of audio
Arabic (Gulf) conversational speech dataset recorded in Saudi Arabia: 500 hours of spontaneous two-person conversations from 50 native speakers, dual-channel, time-aligned human-verified transcripts, signed commercial-use consent
Off-the-shelf list price for a non-exclusive commercial training licence; volume discounts on request. Priced with its MSA set as the dearest of its five Arabic variants (Egyptian, Levantine and Darija are $65), and the vendor's own budget example puts 500 hours each of Egyptian, Gulf and Levantine at $110,000. The full 500 hours lists at $45,000.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
SpeechData.ai catalogue, Arabic (Gulf) conversational speech dataset
$90
per hour of audio
Arabic (Levantine) conversational speech dataset recorded in Lebanon: 500 hours of spontaneous two-person conversations from 50 native speakers, dual-channel, time-aligned human-verified transcripts, signed commercial-use consent
Off-the-shelf list price for a non-exclusive commercial training licence; volume discounts on request. The commercial counterpart to LDC's research-only Levantine telephone sets ($16 to $41 an hour).
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
SpeechData.ai catalogue, Arabic (Levantine) conversational speech dataset
$65
per hour of audio
Nexdata Indonesian spontaneous dialogue smartphone speech, 747 hours from 1,190 native speakers, transcribed with PII redacted (on Datarade)
A 'starts at' one-off licence price, the same $20,000 floor Nexdata shows across its Datarade speech products; spontaneous two-party dialogue at this floor works out cheaper per hour than Nexdata's scripted Indonesian set.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Datarade, Nexdata Indonesian Spontaneous Dialogue Smartphone Speech Dataset, 747 Hours
$20K
per dataset of 747 hours
$27 per hour of audio
Nexdata Indonesian scripted monologue smartphone speech, 359 hours from 496 speakers, transcribed (on Datarade)
A 'starts at' one-off licence price; Nexdata shows the same $20,000 floor on many of its Datarade speech products, so it reads as a minimum order rather than a priced quote. The same corpus lists on ELRA at EUR 57,978.50 for commercial use, about 2.6 times this floor.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Datarade, Nexdata Indonesian Speech Data by Mobile Phone, 359 Hours
$20K
per dataset of 359 hours
$56 per hour of audio
REMI Data licensed native-speaker speech recordings in 13+ languages including Yoruba, Amharic and Swahili, WAV 16 kHz with verbatim transcripts and perpetual commercial consent (on Datarade)
One per-hour rate across all the languages the listing names, Yoruba among them; the stated volume is 100 hours for the whole product, so how many Yoruba hours exist off the shelf is not stated, and custom collection is offered beside it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Datarade, REMI Data Multilingual Speech Recordings, 13+ Languages
$120
per hour of audio
IARPA Babel Vietnamese Language Pack IARPA-babel107b-v0.7 (LDC2017S01): about 201 hours of Vietnamese conversational and scripted telephone speech with transcripts, collected by Appen in 2012
LDC's non-member fee, shown under View Fees (the reduced-licence fee is also $25.00). The Babel packs were paid for by the US government and are priced near zero; non-member LDC licences exclude commercial use, so this is a research price, not a market price for Vietnamese speech.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
LDC catalog, IARPA Babel Vietnamese Language Pack IARPA-babel107b-v0.7 (LDC2017S01)
$25
per corpus of about 201 hours
$0.12 per hour of audio
NaijaVoices suite of speech-text datasets (Hausa, Igbo and Yoruba; 1,800+ hours from 5,000+ speakers): the Supporter tier that adds a waiver for commercial use
A community project's minimum donation for commercial rights, not a negotiated price; the default licence (CC BY-NC-SA 4.0) is free for non-commercial use. Read speech of expert-written sentences, so it sits far below conversational listings.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
NaijaVoices, Membership tiers
$5,000
per commercial-use waiver for the suite of 1,800+ hours
$2.78 per hour of audio
SpeechData.ai Indonesian conversational speech dataset: 2,000 hours of spontaneous two-person conversations recorded in Indonesia, 200 native speakers, dual-channel, time-aligned transcripts
Off-the-shelf asking price for a commercial training and evaluation licence with speaker consent documentation; the page says volume discounts are available for full-dataset and multi-language purchases.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
SpeechData.ai, Indonesian conversational speech dataset
$65
per hour of audio
SpeechData.ai Vietnamese conversational speech dataset: 2,000 hours of spontaneous two-person conversations recorded in Vietnam, 200 native speakers, dual-channel, time-aligned transcripts
Off-the-shelf asking price for a commercial training and evaluation licence with speaker consent documentation; the page says volume discounts are available for full-dataset and multi-language purchases.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
SpeechData.ai, Vietnamese conversational speech dataset
$65
per hour of audio
Real customer-service calls in English and Hindi for chatbot and voicebot training (WiserBrand), usage-based licence
Same per-hour price as the seller's US English calls: no premium for Hindi or code-switched speech at this seller. Withdrawn: on 30 Oct 2026 the listing's address redirected to the provider's Datarade profile (PissedConsumer), which lists no WiserBrand product.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
Datarade, WiserBrand Chatbot Conversation Dataset (English and Hindi)
$25
per hour of audio (usage-based licence)
ALLIES corpus: 900 hours of French broadcast news for diarization evaluation (ELRA catalogue ELRA-S0486)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Broadcast speech in volume, so cheap per hour. A non-commercial-use licence is EUR 2,500 for an academic and EUR 9,375 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, ALLIES Corpus (ELRA-S0486)
$35K
per corpus of 900 hours
Published as 31250.00 € (non-member, commercial use)
$39 per hour of audio
Italian Speech Corpus 1 (Appen): 44 hours from 202 speakers, 4 microphone channels, office and public places (ELRA catalogue ELRA-S0147)
ELRA's non-member price for a company under the commercial-use (VAR) licence. An Appen off-the-shelf corpus distributed through ELRA, one of the few Appen corpora with a public price. A non-commercial-use licence is EUR 1,500 for an academic and EUR 15,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Italian Speech Corpus 1 (Appen) (ELRA-S0147)
$16.8K
per corpus of 44 hours
Published as 15000.00 € (non-member, commercial use)
$382 per hour of audio
Spanish Speech Corpus 1 (Appen): 39 hours from 200 speakers, 4 microphone channels, office and public places (ELRA catalogue ELRA-S0149)
ELRA's non-member price for a company under the commercial-use (VAR) licence. An Appen off-the-shelf corpus distributed through ELRA. A non-commercial-use licence is EUR 1,500 for an academic and EUR 15,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Spanish Speech Corpus 1 (Appen) (ELRA-S0149)
$16.8K
per corpus of 39 hours
Published as 15000.00 € (non-member, commercial use)
$431 per hour of audio
Brazilian Portuguese prompted speech recorded on mobile phones, 1,044 hours from 2,038 speakers (ELRA catalogue ELRA-S0445)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Part of a series priced by language at a fixed rate per hour. ELRA lists no non-commercial licence for it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Brazilian Portuguese Speech Data by Mobile Phone - 1,044 Hours (ELRA-S0445)
$278K
per corpus of 1,044 hours
Published as 247950.00 € (non-member, commercial use)
$266 per hour of audio
BREF-120: 100 hours of French read speech (ELRA catalogue ELRA-S0067)
ELRA's non-member price for a company under the commercial-use (VAR) licence. A classic read-speech corpus. A non-commercial-use licence is EUR 4,000 for an academic and EUR 15,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, BREF-120 - A large corpus of French read speech (ELRA-S0067)
$16.8K
per corpus of 100 hours
Published as 15000.00 € (non-member, commercial use)
$168 per hour of audio
Cantonese Speecon: about 231 hours of adult and child speech over 4 microphone channels in office, car and public places (ELRA catalogue ELRA-S0287)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Multi-microphone, multi-environment recordings. A non-commercial-use licence is EUR 60,000 for an academic and EUR 75,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Cantonese Speecon database (ELRA-S0287)
$84K
per corpus of 231 hours
Published as 75000.00 € (non-member, commercial use)
$364 per hour of audio
Chinese Kids Speech database (Lower Grade): 87 hours from 184 children aged 6 to 10, smartphone (ELRA catalogue ELRA-S0496)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Children's speech. A non-commercial-use licence is EUR 2,600 for an academic and EUR 10,300 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Chinese Kids Speech database (Lower Grade) (ELRA-S0496)
$17.3K
per corpus of 87 hours
Published as 15400.00 € (non-member, commercial use)
$198 per hour of audio
Chinese Kids Speech database (Upper Grade): 72 hours from 161 children aged 10 to 12, smartphone (ELRA catalogue ELRA-S0497)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Children's speech. A non-commercial-use licence is EUR 2,300 for an academic and EUR 9,300 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Chinese Kids Speech database (Upper Grade) (ELRA-S0497)
$15.7K
per corpus of 72 hours
Published as 14000.00 € (non-member, commercial use)
$218 per hour of audio
CLE Pakistan Urdu Speech Corpus: 61.62 hours of phonetically rich read Urdu on varied microphones (ELRA catalogue ELRA-S0403)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Read Urdu here costs several times conversational Urdu in commercial catalogues. A non-commercial-use licence is EUR 15,600 for an academic and EUR 23,400 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, CLE Pakistan Urdu Speech Corpus (ELRA-S0403)
$26.2K
per corpus of 61.62 hours
Published as 23400.00 € (non-member, commercial use)
$426 per hour of audio
Corpus of Spontaneous Japanese (CSJ), 650 hours as listed by ELRA, distributed for NINJAL (ELRA catalogue ELRA-S0488)
ELRA's non-member price for a company under the commercial-use (VAR) licence. The owner, NINJAL, licenses the same corpus directly for 5 million yen (open-ended, up to 10 users), well under ELRA's price. A non-commercial-use licence is EUR 431.25 for an academic and EUR 43,125 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Corpus of Spontaneous Japanese (CSJ) (ELRA-S0488)
$48.3K
per corpus of 650 hours
Published as 43125.00 € (non-member, commercial use)
$74 per hour of audio
Danish SpeechDat(II) FDB-4000: 242 hours of fixed-network telephone speech from 4,000 speakers (ELRA catalogue ELRA-S0073)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Fixed-network telephone speech from 4,000 speakers. A non-commercial-use licence is EUR 48,000 for an academic and EUR 56,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Danish SpeechDat(II) FDB-4000 (ELRA-S0073)
$62.8K
per corpus of 242 hours
Published as 56000.00 € (non-member, commercial use)
$259 per hour of audio
Dutch Polyphone Database: 335 hours of Dutch and Flemish telephone speech (ELRA catalogue ELRA-S0010)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Telephone speech. A non-commercial-use licence is EUR 20,000 for an academic and EUR 35,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Dutch Polyphone Database (ELRA-S0010)
$39.2K
per corpus of 335 hours
Published as 35000.00 € (non-member, commercial use)
$117 per hour of audio
Egyptian Arabic Speecon: 196 hours of Modern Standard Arabic as spoken in Egypt, adults and children, 4 microphone channels in office, car and public places (ELRA catalogue ELRA-S0308)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Multi-microphone, multi-environment recordings. A non-commercial-use licence is EUR 60,000 for an academic and EUR 75,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Egyptian Arabic Speecon database (ELRA-S0308)
$84K
per corpus of 196 hours
Published as 75000.00 € (non-member, commercial use)
$429 per hour of audio
ESTER 2: about 100 hours of transcribed French broadcast news (ELRA catalogue ELRA-S0338)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Commercial use costs several times the non-commercial price. A non-commercial-use licence is EUR 2,000 for an academic and EUR 7,500 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, ESTER 2 Corpus (ELRA-S0338)
$28K
per corpus of 100 hours
Published as 25000.00 € (non-member, commercial use)
$280 per hour of audio
EthioSpeech: 391 hours of read speech in six Ethiopian languages (Amharic, Tigrigna, Oromo, Somali, Afar, Sidama) (ELRA catalogue ELRA-S0494)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Low-resource African languages at a low commercial price per hour. A non-commercial-use licence is free for an academic and EUR 5,400 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, EthioSpeech (ELRA-S0494)
$6,051
per corpus of 391 hours
Published as 5400.00 € (non-member, commercial use)
$15 per hour of audio
French prompted speech recorded on mobile phones, 769 hours from 1,623 speakers (ELRA catalogue ELRA-S0446)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Part of a series priced by language at a fixed rate per hour. ELRA lists no non-commercial licence for it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, French Speech Data by Mobile Phone - 769 Hours (ELRA-S0446)
$205K
per corpus of 769 hours
Published as 182637.50 € (non-member, commercial use)
$266 per hour of audio
French read speech recorded on mobile phones (France, Canada, Africa), 231.9 hours from 406 speakers (ELRA catalogue ELRA-S0465)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Subsets can be requested. ELRA lists no non-commercial licence for it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, French Speech Data by Mobile Phone_Reading - 231 Hours (ELRA-S0465)
$56.8K
per corpus of 231.9 hours
Published as 50670.15 € (non-member, commercial use)
$245 per hour of audio
English spoken by native German speakers (accented English), mobile phones, 535 hours from 1,162 speakers (ELRA catalogue ELRA-S0442)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Accented English priced at the top tier, the same as native German speech in the series. ELRA lists no non-commercial licence for it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, German Speaking English Speech Data by Mobile Phone - 535 Hours (ELRA-S0442)
$171K
per corpus of 535 hours
Published as 152475.00 € (non-member, commercial use)
$319 per hour of audio
German prompted speech recorded on mobile phones (generic, in-car, smart-home scripts), 1,796.70 hours from 3,442 speakers (ELRA catalogue ELRA-S0449)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Top tier of a per-hour price ladder in this series (EUR 285 an hour). ELRA lists no non-commercial licence for it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, German Speech Data by Mobile Phone - 1,796 Hours (ELRA-S0449)
$574K
per corpus of 1,796.70 hours
Published as 512059.50 € (non-member, commercial use)
$319 per hour of audio
German read speech recorded on mobile phones, 211 hours from 327 speakers (ELRA catalogue ELRA-S0473)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Subsets can be requested. ELRA lists no non-commercial licence for it.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, German Speech Data by Mobile Phone_Reading - 211 Hours (ELRA-S0473)
$51.7K
per corpus of 211 hours
Published as 46103.50 € (non-member, commercial use)
$245 per hour of audio
GlobalPhone Hausa: 8.73 hours of read Hausa speech (ELRA catalogue ELRA-S0347)
ELRA's non-member price for a company under the commercial-use (VAR) licence. A small research corpus; per hour it is far dearer than larger commercial Hausa sets. A non-commercial-use licence is EUR 700 for an academic and EUR 3,600 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, GlobalPhone Hausa (ELRA-S0347)
$4,034
per corpus of 8.73 hours
Published as 3600.00 € (non-member, commercial use)
$462 per hour of audio
GlobalPhone Tamil: 20 hours of read Tamil speech (ELRA catalogue ELRA-S0205)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Other GlobalPhone languages list at EUR 3,600 for about 20 hours, six times this price. A non-commercial-use licence is EUR 125 for an academic and EUR 600 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, GlobalPhone Tamil (ELRA-S0205)
$672
per corpus of 20 hours
Published as 600.00 € (non-member, commercial use)
$34 per hour of audio
Gram Vaani: 130 hours of Hindi phone recordings from 4,000 mostly rural speakers in Bihar, Jharkhand and Madhya Pradesh (ELRA catalogue ELRA-S0405)
ELRA's non-member price for a company under the commercial-use (VAR) licence. Rural, accented, real-world telephone speech from a community voice platform. A non-commercial-use licence is free for an academic and EUR 50,000 for a company.
Audio and speechAsking prices off the shelfOct 2026A catalogue or marketplace listing
ELRA catalogue, Gram Vaani data set (ELRA-S0405)
$56K
per corpus of 130 hours
Published as 50000.00 € (non-member, commercial use)
$431 per hour of audio

Where this data comes from, and who lists it

The whole supply map

41 hours
Ramsa is a developing 41-hour Emirati Arabic corpus from 157 speakers covering Urban, Bedouin and Mountain (Shihhi) subdialects, built for research; the only priced Emirati speech found is ELRA's 85.65-hour read UAE corpus (EUR 36,000).
Reaches buyers: Academic research corpus, not offered for sale
arXiv 2603.08125, Ramsa: A Large Sociolinguistically Rich Emirati Arabic Speech Corpus for ASR and TTS, 9 Mar 2026
668 hours
SADA holds 668 hours of audio from Saudi television shows across several dialects and environments; it circulates free for research, and SpeechData.ai's licence table lists it as non-commercial.
Reaches buyers: Free research release; not licensed for commercial training
arXiv 2508.12968, Arabic ASR on the SADA Large-Scale Arabic Speech Corpus with Transformer-Based Models, 18 Aug 2025

Raises the price

Languages and dialects with little recorded speech
Natural conversation rather than read text
Transcripts in the language, with speaker labels
Thousands of hours from many speakers

The rights question

The same as any recording: what the people speaking were told, and whether the recordings are the company's or a client's. Recordings made in another country follow that country's consent and privacy law as well as the buyer's.

The guide on what else to settle