Order from Lab 7B362A

Low resource languages

Native-speaker recordings with human transcripts in under-served languages (Wolof, Fula, Kinyarwanda, Quechua and others), each clip tied to a speaker record with dialect, native language and a written consent covering commercial AI training. Alongside it, human-translated parallel sentence pairs for the same languages, with who translated and by what method.

How answering works

Any number of people fill one order. The lab receives one dataset, in its own columns.

Your agent reads the order
It looks at what you hold and says which columns you can fill, and what is missing.
Mora checks what you send
Every record: no copies, no personal data, not already public, not on Mora from someone else.
The lab pays per record it accepts
The lab named the price. You fill as much of the order as you can.

The columns the lab wants

You do not need every column. Mora says which ones each dataset fills.

Column Type What it must hold

clip_id id One recording

audio audio FLAC or WAV, 16 kHz or better, mono, never re-encoded from MP3

duration_s number Length of the clip in seconds

language label ISO 639-3 code of the language spoken

dialect label Region and Glottocode, declared by the speaker

speaker_id id One real person, the same id on every clip of theirs

native_language label The speaker's first language, to tell native from second-language speakers

gender label Speaker gender, self-declared

age_band label Speaker age range

consent_id id Link to the consent record: form version, date signed, form language, scope (AI training, commercial use, resale), paid or not, how to withdraw

speech_type label Read or spontaneous

prompt_id id For read speech, the sentence that was read

transcript text Verbatim text in the language's own script, code-switching marked

transcriber_id id Who transcribed

reviewer_id id The second native speaker who checked the transcript

device label Recording device and setting

pair_id id One parallel sentence pair

src_text text Source sentence

src_lang label ISO 639-3 code of the source

tgt_text text Translated sentence

tgt_lang label ISO 639-3 code of the target

original_side label Which side was written first

translator_id id Who translated

method label Human, post-edited, or machine

domain label News, health, religious, conversation and so on

Column	Type	What it must hold
`clip_id`	id	One recording
`audio`	audio	FLAC or WAV, 16 kHz or better, mono, never re-encoded from MP3
`duration_s`	number	Length of the clip in seconds
`language`	label	ISO 639-3 code of the language spoken
`dialect`	label	Region and Glottocode, declared by the speaker
`speaker_id`	id	One real person, the same id on every clip of theirs
`native_language`	label	The speaker's first language, to tell native from second-language speakers
`gender`	label	Speaker gender, self-declared
`age_band`	label	Speaker age range
`consent_id`	id	Link to the consent record: form version, date signed, form language, scope (AI training, commercial use, resale), paid or not, how to withdraw
`speech_type`	label	Read or spontaneous
`prompt_id`	id	For read speech, the sentence that was read
`transcript`	text	Verbatim text in the language's own script, code-switching marked
`transcriber_id`	id	Who transcribed
`reviewer_id`	id	The second native speaker who checked the transcript
`device`	label	Recording device and setting
`pair_id`	id	One parallel sentence pair
`src_text`	text	Source sentence
`src_lang`	label	ISO 639-3 code of the source
`tgt_text`	text	Translated sentence
`tgt_lang`	label	ISO 639-3 code of the target
`original_side`	label	Which side was written first
`translator_id`	id	Who translated
`method`	label	Human, post-edited, or machine
`domain`	label	News, health, religious, conversation and so on

How much, in the lab's words

2,000 transcribed hours of speech (200 hours per language across 10 languages, at least 300 native speakers per language) and 1 million human-translated sentence pairs (100,000 per language pair)

recordings
speakers
consent
transcripts
parallel-text
prompts