Translator / interpreter
UZ–RU–EN pair, CAT, a niche — plain text got cheaper, liability did not.
Income range: 4,000,000–25,000,000 UZS / mo
Takeaways
- Written translation and interpreting are two jobs. CAT tools in writing; in the booth, voice and 20–30 minute turns.
- There is no official salary series. UZ–RU–EN is the common pair; legal, medical, and technical work pays more. AI cheapens plain text, not liability.
- “I speak two languages” is not enough. Native-language editing, terms, and notary/apostille rules are the filter.
Moving meaning — not words
A translator moves written text and edits it. An interpreter moves speech now. A localizer fits a product to a culture (units, jokes, buttons). Escort interpreting is travel and meetings.
Uzbekistan streams: notary/apostille papers, business contracts, state events, content/subtitles, IT UI. “I translate everything” is cheap and unsafe.
A fit if you care about words, can sit quietly (writing) or hold a voice (speech). A poor fit if “I roughly got it” is your standard and deadlines slide forever.
Modes — do not mix them
| Mode | What you do | The body |
|---|---|---|
| Written | Documents, web, subs, CAT | Eyes, back, a deadline |
| Consecutive | Speech — pause — render | Notes, voice, a meeting |
| Simultaneous | Booth pair, 20–30 minutes | Fatigue, water, ears |
| Whispered / escort | Beside one person | Travel, broken hours |
| Localization | UI, games, marketing | Glossary, context, QA |
A writing day and a booth day
Writing: brief (tone, deadline, format), glossary, CAT (Trados, memoQ, OmegaT), edit, QA (numbers, names, tags). Hours of screen — breaks.
Speech: pre-read at night, water, turns with a partner, how to say “I do not know.” After 20–30 minutes errors rise — that is physiology, not weakness.
Mixing a morning document and an afternoon event is burnout. Make one mode primary.
Pairs and niches
UZ–RU–EN is the common ad and order. An extra language (German, Korean, Chinese, Turkish) can open a door — but if native editing is weak, language count will not save you.
Niches: law (contracts, court), medicine (notes, consults), tech/IT, finance. Generic “news translation” got cheaper with AI. Learning a niche means reading the field, not only a dictionary.
Notary and apostille — this page is not a stamp
Path: language, niche, CAT
01Native editing
Clean prose in Uzbek or Russian. Read news and statute language, not only shows.
02Second language near C1
Speaking is not enough. Writing and listening. IELTS is optional; a work sample matters more.
03CAT and three samples
OmegaT is free. Three anonymized samples (law or tech). ProZ/Telegram with care.
04First paid job
A small volume, a deposit, a date. Cap unpaid “for experience.”
Tools
- CAT: OmegaT (free), then Trados/memoQ at the client
- Glossary and TM — the same term twice
- QA: numbers, tags, brackets
- Interpreting: headset, a note system, water
Who writes “I do not know”
A fit: quiet precision, deadlines, writing “I do not know.” For interpreting: a voice and calm in front of people. A poor fit: guessing and hiding lateness. Eyes and ears without a routine are a risk.
Mistakes that lose the client
- Taking every niche at one price
- Pasting a secret paper into ChatGPT
- Holding a booth alone for two hours
- Taking notary work on a blogger’s book
- Delivering raw machine output
Pay: rates exist, a series does not
There is no official “average translator salary” — we will not invent one. In practice: word, page, hour, or project. Niche and rush add a premium. In-house wages (agency, IT, state) vary in ads — open 5–10. Tashkent’s ~12.01M (H1 2026) is context, not a profession row.
AI post-edit is usually cheaper than a full translation. Name it in the contract. A foreign agency: USD/EUR, an NDA, sometimes ISO — that is not an “official UZ median” either.
Where the work comes from
Referrals, law firms, notaries, IT/outsource, state events, Telegram, marketplaces with care. A deposit and a volume. Language: source Uzbek/Russian, target English — or the reverse.
University translation/philology helps, and is not required for content. Notary tracks often ask for the diploma.
Agencies, ISO, relocation
A foreign agency wants written English, CAT, sometimes a test. Relocation is a visa and that country’s register — not a how-to here. Booth work in Europe is tied to a pair and labor rules. Do not convert a rate at the FX and say “me too.”
Machine translation and post-edit
DeepL, Google, and LLMs cheapened plain text. Post-edit work grows — a medical/legal error is still yours. AI does not yet own simultaneous booths (noise, speech, liability).
Learn: when MT is allowed, when it is not, secrecy, and telling the client “this is machine.”
What remains
Generic text will shrink. Niche, interpreting, notary, and liability remain. Risks: eyes, voice, a solo booth, a broken NDA. Myths: “two languages = a career”; “AI replaces everyone.”
Beside the translation desk
Is a degree required?
For notary and some tenders, yes. For content/UI a portfolio can suffice. Read the ad.
What do you charge per word or page?
No official average. Word/page/hour. Compare 5–10 offers. Post-edit is cheaper.
Will AI take the work?
It will shrink cheap general text. Niche, interpreting, and liability remain.
Should I do both writing and interpreting?
One at the start. Interpreting is voice; writing is eyes. Both at once burns you out.
Where are the jobs?
Referrals, agencies, Telegram, marketplaces with care. A deposit and a contract.
Which pair should I pick?
Your native language plus what the market asks (often EN or RU). Do not switch every month.
Is CAT required?
Often in agencies and IT. A small notary page may still be Word. OmegaT is a free trial.
What do I do in week one?
Edit one page in your native language, open OmegaT, read one niche text, write a rate.
Figures draw on 2026 sources: Uzbekistan National Statistics Committee (official sector averages, Q1/H1 2026), Flexa and hh.uz vacancy medians, and public IT Park / product-team guides. Not a guarantee — contract type, city, tax, and level change the range. Compare live vacancies yourself.