AI Data · Capacity

Asian-language capacity,
locale by locale

Which Asian locales we can staff for speech data and transcription, what each one has delivered, and which are only ready to activate. Capacity here means people we can field, not the size of a database.

Ask about a locale →
How to read this page
Two statuses, kept apart

Production-proven

The locale has delivered production batches on a live AI data program in 2026, with a fielded team, a written locale guide and QA records.

Ready to activate

Recruitment channels and screening are prepared, but no production batch has been delivered yet. Every one of these starts with a pilot batch on your specification.

Production-proven
Seven locales with delivered work

Cantonese · yue-HK

Transcription validation; Cantonese and Cantonese–English recording, including hundreds of two-speaker service-call scripts on separated tracks; on-device testing in Hong Kong. Speakers are Hong Kong-based. The recording line went from pilot to hundreds of scripts within three weeks.

Taiwan Mandarin · zh-TW

Transcription validation including Mandarin–English switching, on a Traditional-character guide where every file passes an automated Simplified-character check. A Taiwan collection and transcription pool is recruiting now, including speakers aged 55 and over.

Mandarin, Mainland · zh-CN

Transcription validation and Mandarin–English code-switching, in weekly batches.

Korean · ko-KR

Transcription validation and Korean–English code-switching; proofreading and quality scoring of Korean–Chinese song-lyric data.

Japanese · ja-JP

Transcription validation and Japanese–English code-switching.

Tagalog · tl-PH

Transcription validation and Tagalog–English (Taglish) code-switching.

Turkish · tr-TR

Transcription validation and Turkish–English code-switching. Not an Asian language, but it runs on the same line and the same QA.

Ready to activate
Seven code-switching pairs, pilot first

Hindi–English (Hinglish) · Malay–English · Tamil–English · Indonesian–English · Thai–English · Vietnamese–English · Gulf Arabic–English

If a locale is not on this page, assume we do not have a fielded team for it. Ask, and we will tell you plainly rather than start recruiting on your deadline.

What sets capacity
Why the number of people is not the number that matters

Screening, not headcount

Filtering a pool on declared native language returns people who speak the language, not people who switch the way your users do. We screen for demonstrated everyday switching. On one seven-language program, time per task in most languages fell below half of the first month’s by the second month.

Fixed core, open intake

A small team stays on each locale between batches. New people come in through locale-specific qualification tests on our own task platform, so volume can rise without restarting onboarding.

Hard demographics take lead time

Older speakers and narrow regional accents are the slowest to recruit. For Taiwan we reach speakers aged 55 and over through community channels and family referrals, and quote quotas only after a pilot.

Asking for capacity
Send five things, get a firm-or-estimate answer

What to send

Locale and variant · task type (collection, transcription, validation, scoring) · hours or units per week · speaker quotas · start date and turnaround.

What comes back

Per locale: people we can field, weekly throughput, each figure marked firm or estimate, and the risks we see. Rows we cannot staff stay blank.

Then a pilot

A small batch on your specification, reviewed by you before any volume is agreed. See how the data line works or the notes for vendor managers.

Common questions
Capacity,
answered

Which Asian languages can you staff for speech data today?

Seven locales have delivered production work: Cantonese (Hong Kong), Taiwan Mandarin, Mainland Mandarin, Korean, Japanese and Tagalog, plus Turkish on the same line, each including English code-switching. Hindi, Malay, Tamil, Indonesian, Thai, Vietnamese and Gulf Arabic paired with English are ready to activate with a pilot.

What does ready to activate mean?

Recruitment channels and screening are prepared but no production batch has been delivered in that locale yet. We start with a pilot batch on your specification and only commit to volume after it passes your review.

Is capacity a headcount or a database size?

Headcount we can actually field for your task and turnaround. A large database of registered names says little about how many of them pass a screen for natural code-switching or a specific demographic.

Can you recruit older speakers or specific accents?

Yes, with lead time. We are building a Taiwan Mandarin pool that includes speakers aged 55 and over, reached through community channels and family referrals. Quotas are confirmed after a pilot, not before.

Need a locale staffed?

Tell us the locale, task, weekly volume and speaker quotas — we'll answer row by row.

Ask about capacity →