Skip to main content
CPContinental PeopleHuman visual data for AI

AUDIO-ONLY · MULTILINGUAL SPEECH COLLECTION

From a multilingual brief
to real speech samples.

We coordinate speaker recruitment, recording tasks, audio collection and contributor payments, so your team can focus on evaluating the speech it needs.

Listen to a real sample Explore the work
Continental PeopleOctober 2026Audio, recorded by real people
RECRUITMENT IN MOTION
285.

unique applications processed
in three days

07 OCT · LAUNCH10 OCT · 285 PROCESSED

A 12-language recruitment brief.
Real recordings from paid contributor trials.

A language list tells you what to collect.
It does not yet tell you how to get it recorded.

For an audio-only brief spanning 12 languages, we organised remote recruitment and short reading tasks. Within three days, we processed 285 unique applications. The work went on to produce real speech samples and paid contributor trials.

This single brief covered languages used across Eastern Europe, the Caucasus and Central Asia.

Short recordings make a collection brief tangible: your team can hear the voices and assess the reading task before planning the next batch.

01

START WITH THE COLLECTION DESIGN

What sits behind
a list of languages?

“We need speech in these languages” is a starting point. The number of voices, the reading task and the amount required from each person turn it into a collection plan.

THE RECRUITMENT BRIEF12 languages in the recruitment brief
01Armenian
02Azerbaijani
03Bashkir
04Belarusian
05Georgian
06Kazakh
07Kyrgyz
08Romanian
09Tajik
10Tatar Audio example ↓
11Turkmen
12Uzbek

The speaking task

Reading a supplied script, spontaneous speech or a conversation require different instructions and checks. For this collection, we used short scripted readings.

The voices

Language, reading confidence and any dialect or participant requirements belong in the brief before the first invitation goes out.

The recording

The device, room, file format and target duration shape both the participant’s task and the review process.

02

MAKE THE SEARCH ACTIONABLE

From interest
to a recording task.

Fast recruitment works best when each next step is clear. We asked applicants about their language and reading confidence, available equipment, a quiet recording space and weekly availability.

Selected applicants received a reading text and clear instructions for recording and submitting their audio. Each invitation built on the details the applicant had already shared.

  1. 07 OCT

    Open the search

    A remote recording opportunity with a clear language list and paid trials.

  2. 08 OCT

    Make it personal

    Clarify missing details and send selected applicants their recording instructions.

  3. 09 OCT

    Collect real files

    Trial recordings collected from nine recruited contributors.

  4. 10 OCT

    285 unique applicants

    All nine contributors who submitted a paid trial received payment.

03

LISTEN TO THE MATERIAL

A real voice.
A tangible first result.

The Tatar sample makes the work concrete: a person reading a supplied everyday story. Explore the opening or listen to the longer recording.

TATAR · SCRIPTED READING

The old chair’s new day

A short everyday story, read by a real speaker.

0:000:30

THE OPENING PASSAGE

On Saturday morning I woke up a little early. The house was quiet, and through the window I could see the trees in the courtyard. Today there was no need to hurry anywhere. I put the kettle on and remembered a small task I had been putting off.

The opening passage of the reading text.

TaskScripted reading
LanguageTatar
Full recording4 min 46 sec
04

FROM TASK TO RECORDING

Your team evaluates the audio.
We coordinate the collection.

We handle participant instructions, submissions and payments, keeping the collection moving. Follow one paid Uzbek recording trial from the reading task to the recorded sample.

A PAID RECORDING TRIALUZBEK

The work produced a documented candidate pool and real speech samples, supported by a practical workflow for remote recording, file collection and contributor payments.

05

DESIGN YOUR FIRST BATCH

Start with the voices.
Then plan the hours.

The same 100 hours can come from ten people or a hundred. That choice changes recruitment, the number of individual recording tasks and the diversity of voices.

Choose the audio volume and hours per speaker to see how many voices your collection needs. Then define the languages, voice profiles, scripts and quality requirements for a first batch.

EXPLORE A COLLECTION

More voices can bring greater variety. More hours per voice can bring greater depth.

MINIMUM VOICES TO PLAN FOR
20

100 audio hours ÷ 5 hours per voice = 20 voices

At 1 hour / voice100 voices
At 10 hours / voice10 voices

Start with your language list.
These details help shape the collection.

  1. Languages and any dialect requirements
  2. Number and profile of unique voices
  3. Target duration per language and per person
  4. A representative script or recording task
  5. File specifications and quality requirements

CUSTOM SPEECH DATA COLLECTION

Discuss an audio-only
speech collection.

Share your language list and recording goal. We’ll help shape the speaker profiles, sample task and a practical first batch.

Discuss a speech-data pilot