Armenian Transliterator Get API key

Tickets arrive in Latin letters.
Agents read Armenian.

Customers write to chat and email the way they type every day — in Latin letters. An agent can usually work out what a message means, word by word. But a helpdesk full of romanized text is slow to read, impossible to search properly, and awkward to quote in a reply. Convert the message on the way in, and the whole queue speaks Armenian.

  1. 1The problemRomanized tickets are slow to read, and search can’t index them.
  2. 2The requestConvert each message once, as it enters your helpdesk.
  3. 3The responseInstant Armenian, plus a polished version in a formal tone.
  4. 4The resultFaster handling, and an archive that search actually finds.
The problem

A queue that reads like a chat app

The ticket below is a normal return request. Everything in it is Armenian — except the script it’s typed in.

What arrives from live chat

Customer 12:41

Barev dzez, gnel em payusak, isk chisht guyn@ chi. Khndrum em poxel.

Agent 12:44

Barev, khndrum em greq hramani hamar@.

Stored like this, the ticket is hard to scan — and a later search for պայուսակ (“bag”) won’t find it.

Why it costs you

  • Reading speed drops. Agents decode romanized text word by word. Across a shift, that’s minutes per ticket and more misreadings on a bad one.
  • The archive stays dumb. Keyword search, tags and macros work on Armenian script. Romanized history is effectively unsearchable.
  • Quoting is embarrassing. Pasting “guyn@ chi” into a formal reply, a report, or a refund note looks unprofessional in front of the customer.
The integration

One call when the message arrives

Send the customer’s text to the API as it enters your helpdesk — from chat, email or a web form — and store what comes back. Ask for the polished version in a formal tone; it’s what your agents should read and your archive should keep.

Request · POST /api/v1/transliterate
curl -X POST https://armtranslit.com/api/v1/transliterate \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Barev dzez, gnel em payusak, isk chisht guyn@ chi. Khndrum em poxel.", "tone": "formal"}'
Response · 200 OK
{
  "id": "019b6a42-7b14-7e03-9c6d-41f8b2e5a729",
  "input": "Barev dzez, gnel em payusak, isk chisht guyn@ chi. Khndrum em poxel.",
  "raw_transliteration": "Բարև ձեզ, գնել եմ պայուսակ, իսկ ճիշտ գույն@ չի։ Խնդրում եմ փոխել։",
  "refined_text": "Բարև ձեզ։ Գնել եմ պայուսակ, իսկ ճիշտ գույնը չի։ Խնդրում եմ փոխել։",
  "tone": "formal",
  "used_llm": true,
  "matched_words": 7,
  "total_words": 12
}
In your ticket intake · Python
r = httpx.post(
    "https://armtranslit.com/api/v1/transliterate",
    headers={"X-API-Key": "YOUR_API_KEY"},
    json={"text": message, "tone": "formal"},
)
body = r.json()
ticket.body_hy = body["refined_text"] or body["raw_transliteration"]

Instant vs. polished

raw_transliteration
The instant conversion, always returned. Note it keeps the customer’s own shorthand — guyn@ stays գույն@.
refined_text
The same message with grammar and punctuation cleaned up: գույն@ becomes գույնը, sentences are capitalised and closed properly. This is the text agents read.
tone: "formal"
Sets the register of the polished version — the right default for customer-facing queues. auto follows the customer’s own tone.
id
If a word ever comes out wrong, send the fix to /api/v1/corrections with this ID. Reviewed fixes improve everyone’s results.
The business result

The same ticket, converted on arrival

The agent sees proper Armenian with correct punctuation. The archive stores the same thing, so every past ticket becomes searchable.

What the agent reads now

Customer 12:41 · converted

Բարև ձեզ։ Գնել եմ պայուսակ, իսկ ճիշտ գույնը չի։ Խնդրում եմ փոխել։

Agent 12:43

Բարև ձեզ։ Ներեցեք անհարմարության համար․ կուղարկենք փոխարինող պայուսակ այսօր։

“Hello, I bought a bag but the colour isn’t right. Please replace it.” — handled in two minutes, in one script.

  • Agents read at full speed. Proper script, native punctuation, consistent tone — no decoding, fewer misreadings.
  • The archive becomes an asset. Tickets stored in Armenian script are found by keyword search, feed reporting, and make macros match.
  • Mistakes get fixed once, for everyone. A wrong word becomes a correction with the request ID; reviewed fixes land in the shared dictionary.
  • Customer data gets the care it needs. Requests are deleted after 30 days, never train AI, and no-log mode stores nothing at all. The full privacy terms
FAQ

Support team questions

Anything else? Write to us and we’ll help.

Do you train on our customers’ messages?

No. Requests are deleted after 30 days, and neither your text nor your corrections are used to train AI models. Business customers can switch their key to no-log mode, where nothing is written to the database at all. See Trust & privacy.

Our data-protection review will ask where the text goes. What do we tell them?

The instant result is produced on our own servers. The polished version sends the text to OpenAI or Anthropic through their API, and you can skip that step entirely with use_llm: false. We answer security-questionnaire questions directly — get in touch.

Does it work for live chat as well as email tickets?

Yes — both are just text. For live chat you may prefer the instant result only (use_llm: false) for the fastest round-trip, and the polished version for email, where a few seconds don’t matter.

More use cases

See it in another context

E-commerce searchLatin-letter queries find Armenian-script products. AI & data normalizationOne Armenian form for every Latin spelling, before analytics or training.

Give your agents the message, not the puzzle

Convert tickets and chat on arrival. Get a key, run your worst ticket through the playground, and see what your queue could read like.

Get an API key Try the playground