Why We Trained 0.8B and 4B Transliteration Models for Mixed-Script Input Methods (No Mode-Switching)

losdwind — Sat, 07 Nov 2026 09:00:13 +0000

If you write software documentation, Slack messages, or bilingual notes across two scripts—say Simplified Chinese + English, Japanese + English, Hindi + English, or Arabic + English—you already know the mode-switching tax of traditional input method editors (IMEs).

Every time an English identifier (README, useEffect, roadmap, v1.0.22) appears in the middle of a Pinyin or Romaji sentence, a conventional character-by-character IME forces you to toggle layouts (Ctrl+Space or the Globe key), pick candidates, toggle back for the English token, and toggle again.

On the other hand, general-purpose LLM translators fail in the opposite direction: if you feed them a mixed draft like jintian yao finish README or kyou wa Tokyo de meeting ga arimasu, they translate the entire sentence into a single language instead of preserving intentional bilingual code-switching.

That gap is why we built UnimeType—a cross-platform input method and set of browser converters—and open-sourced our specialized Transliteration 2.0 (0.8B & 4B) models under Apache-2.0.

1. The Core Problem: Transliteration != Translation

In a mixed-script input workflow, the user types phonetic Latin text on a single QWERTY layout:

Chinese (Pinyin + English): women xiazhou release v2.1, qing xian check PR → 我们下周 release v2.1，请先 check PR
Japanese (Romaji + English): kyou wa Tokyo de team meeting ga arimasu → 今日は Tokyo で team meeting があります
Hindi (Hinglish + English): kal morning mein deployment check karna hai → कल morning में deployment check करना है
Arabic (Arabizi + English): ana fil meeting dilwa2ti, hab3at el PR ba3dein → أنا في ال meeting دلوقتي، هبعت ال PR بعدين
Russian (Translit + English): davai zavtra obsudim release plan → давай завтра обсудим release plan

A model serving an input method has three hard constraints that general chat LLMs violate out of the box:

Boundary preservation: Recognized English words, code identifiers, flags, URLs, and @handles must stay in Latin script unless they are phonetic syllables of the target script.
Ambiguous homophone disambiguation in context: Token-level syllables like you (English pronoun you vs. Chinese Pinyin 有/又), main (English main vs. Hindi मैं), or de / to (English prepositions vs. Chinese/Japanese particles) can only be resolved by looking at the surrounding clause.
Sub-second latency & zero conversational chatter: An IME cannot wait 3 seconds for a chain-of-thought reasoning block or strip conversational filler like "Here is your converted text:".

2. Training `UnimeType/Transliteration-2.0-0.8B` and `4B`

To make full-sentence mixed-script conversion fast enough for local desktop inference (via LM Studio / llama.cpp / MLX) and low-latency cloud routing, we fine-tuned two model sizes on Qwen/Qwen3.5:

UnimeType/Transliteration-2.0-0.8B (GGUF · Safetensors): Ultra-lightweight model (~500–800 MB quantized) designed for instant local CPU/GPU inference on laptops.
UnimeType/Transliteration-2.0-4B (GGUF · Safetensors): High-accuracy 4B model for complex code-switched sentences, regional dialects (Egyptian/Levantine/Gulf Arabizi), and polyphonic kanji selection.

Both models use a compact Short-Protocol (v2.1) system prompt with thinking disabled (enable_thinking=False) so the first output token is emitted immediately.

We published our full benchmark methodology—covering 900+ multi-script test cases, homophone collision traps, and English-retention stress tests—in How We Evaluate Transliteration Models.

3. Try It in the Browser (Zero Install) or Locally

We recently shipped four free, no-sign-in browser converters powered by the same conversion pipeline so you can test full-sentence mixed-script conversion directly in the browser:

Free Pinyin to Chinese Sentence Converter (see also the Latin to Chinese Keyboard guide)
Free Romaji to Hiragana & Kanji Converter (see also the Japanese Romaji Keyboard guide)
Free Hinglish (English to Hindi Devanagari) Converter (see also the Latin to Hindi Keyboard guide)
Free Russian Translit to Cyrillic Converter (and the English to Arabic Arabizi Keyboard guide)

If you want system-wide conversion inside any native app (VS Code, Slack, Notion, WeChat, browsers, terminals), along with in-place AI Polish Writing and AI Text Explainer:

Download native clients (macOS, Windows, iOS, Android, Linux, ChromeOS): https://unimetype.com/download
macOS Homebrew Tap: brew tap losdwind/tap && brew install --cask unimetype
Local offline inference guide (LM Studio + GGUF): AI Service Models & Local Setup Docs

Feedback on tricky bilingual boundary cases or host-editor quirks is very welcome!

DEV Community: losdwind

Why We Trained 0.8B and 4B Transliteration Models for Mixed-Script Input Methods (No Mode-Switching)

1. The Core Problem: Transliteration != Translation

2. Training UnimeType/Transliteration-2.0-0.8B and 4B

3. Try It in the Browser (Zero Install) or Locally

2. Training `UnimeType/Transliteration-2.0-0.8B` and `4B`