Translation engines
Choose from three translation engines. All outputs for the same sentences between English and 28 languages are available. Read the outputs for your languages before choosing.
Choosing an engine
| Engine | Runs on | Setup | Text destination |
|---|---|---|---|
| Azure Translator | Cloud | Your Azure key (setup); 2 million characters/month free | Microsoft servers |
| PLaMo(PLaMo 3 Translate 0.8B) | This PC | Download ~0.57GB model | Stays local |
| TranslateGemma(TranslateGemma 4B) | This PC | Download ~2.49GB model | Stays local |
- Set first-to-third priorities in Settings > Translation. The first engine supporting the language pair and ready to use (key entered or model downloaded) is selected.
- PLaMo supports a limited set of languages; it is not used for the extremely low-scoring languages below. Unsupported pairs pass to the next engine.
- Local engines run on CPU or NVIDIA GPU. GPU is faster but shares VRAM with VRChat and speech synthesis.
Suggested reading order:
- Select a language below and open both English → that language and that language → English.
- Check VRChat terms (instance, avatar), negation, times, numbers and refusals.
- Do not choose by score alone. Scores measure character similarity to one reference, not the percentage of correct meaning.
Overview
Cells show “non-empty outputs/30 · chrF.” Click a language to see 30 source sentences, references and all three engines’ actual outputs.
| Language / full output | Azure EN→X | PLaMo EN→X | Gemma EN→X | Azure X→EN | PLaMo X→EN | Gemma X→EN |
|---|---|---|---|---|---|---|
| Japanese | 30/30 · 53.3 | 30/30 · 56.7 | 30/30 · 43.3 | 30/30 · 69.7 | 30/30 · 77.6 | 30/30 · 64.6 |
| Simplified Chinese | 30/30 · 61.2 | 30/30 · 63.2 | 30/30 · 45.7 | 30/30 · 75.0 | 30/30 · 80.5 | 30/30 · 69.8 |
| Traditional Chinese | 30/30 · 61.8 | 30/30 · 58.4 | 30/30 · 46.2 | 30/30 · 75.4 | 30/30 · 79.9 | 30/30 · 71.5 |
| Korean | 30/30 · 43.6 | 30/30 · 44.6 | 30/30 · 38.5 | 30/30 · 77.3 | 30/30 · 76.6 | 30/30 · 65.5 |
| Spanish | 30/30 · 84.2 | 30/30 · 81.9 | 30/30 · 74.2 | 30/30 · 88.6 | 30/30 · 87.5 | 30/30 · 80.9 |
| French | 30/30 · 82.6 | 30/30 · 71.8 | 30/30 · 65.7 | 30/30 · 82.6 | 30/30 · 79.6 | 30/30 · 80.6 |
| German | 30/30 · 85.3 | 30/30 · 77.3 | 30/30 · 62.3 | 30/30 · 85.5 | 30/30 · 82.7 | 30/30 · 80.4 |
| Portuguese | 30/30 · 88.1 | 30/30 · 15.2 | 30/30 · 77.6 | 30/30 · 89.3 | 30/30 · 50.9 | 30/30 · 81.3 |
| Russian | 30/30 · 76.0 | 30/30 · 67.5 | 30/30 · 61.9 | 30/30 · 77.5 | 30/30 · 78.7 | 30/30 · 75.9 |
| Italian | 30/30 · 83.5 | 30/30 · 81.8 | 30/30 · 69.4 | 30/30 · 85.9 | 30/30 · 84.1 | 30/30 · 81.0 |
| Vietnamese | 30/30 · 69.4 | 30/30 · 62.9 | 30/30 · 55.9 | 30/30 · 76.5 | 30/30 · 74.2 | 30/30 · 70.0 |
| Indonesian | 30/30 · 79.2 | 30/30 · 72.3 | 30/30 · 67.1 | 30/30 · 77.4 | 30/30 · 76.8 | 30/30 · 76.1 |
| Turkish | 30/30 · 83.2 | 30/30 · 3.1 | 30/30 · 63.5 | 30/30 · 85.7 | 30/30 · 35.0 | 30/30 · 72.3 |
| Polish | 30/30 · 83.7 | 30/30 · 72.8 | 30/30 · 64.7 | 30/30 · 81.8 | 30/30 · 83.6 | 30/30 · 74.1 |
| Dutch | 30/30 · 85.3 | 30/30 · 81.1 | 30/30 · 72.3 | 30/30 · 85.6 | 30/30 · 86.5 | 30/30 · 80.0 |
| Swedish | 30/30 · 91.9 | 30/30 · 15.6 | 30/30 · 77.4 | 30/30 · 85.6 | 30/30 · 23.5 | 30/30 · 79.8 |
| Danish | 30/30 · 83.8 | 30/30 · 1.8 | 30/30 · 71.2 | 30/30 · 83.3 | 30/30 · 16.7 | 30/30 · 80.2 |
| Finnish | 30/30 · 82.3 | 30/30 · 11.5 | 30/30 · 67.2 | 30/30 · 88.5 | 30/30 · 20.4 | 30/30 · 80.0 |
| Czech | 30/30 · 72.1 | 30/30 · 11.0 | 30/30 · 50.5 | 30/30 · 86.1 | 30/30 · 49.1 | 30/30 · 74.5 |
| Greek | 30/30 · 72.0 | 30/30 · 2.2 | 30/30 · 56.2 | 30/30 · 77.4 | 30/30 · 13.2 | 30/30 · 79.0 |
| Hungarian | 30/30 · 74.4 | 30/30 · 70.9 | 30/30 · 49.9 | 30/30 · 83.2 | 30/30 · 78.8 | 30/30 · 72.3 |
| Romanian | 30/30 · 75.2 | 30/30 · 3.0 | 30/30 · 62.5 | 30/30 · 80.7 | 30/30 · 20.7 | 30/30 · 78.7 |
| Ukrainian | 30/30 · 71.6 | 30/30 · 18.8 | 30/30 · 52.4 | 30/30 · 76.0 | 30/30 · 50.5 | 30/30 · 73.5 |
| Malay | 30/30 · 77.1 | 30/30 · 72.6 | 30/30 · 67.2 | 30/30 · 80.3 | 30/30 · 81.2 | 30/30 · 72.6 |
| Thai | 30/30 · 77.2 | 30/30 · 69.8 | 30/30 · 57.8 | 30/30 · 74.4 | 30/30 · 73.5 | 30/30 · 71.6 |
| Arabic | 30/30 · 70.9 | 30/30 · 66.1 | 30/30 · 60.7 | 30/30 · 85.4 | 30/30 · 78.1 | 30/30 · 76.3 |
| Hindi | 30/30 · 67.1 | 30/30 · 67.1 | 30/30 · 64.3 | 30/30 · 83.8 | 30/30 · 79.0 | 30/30 · 68.7 |
| Filipino (Tagalog) | 30/30 · 81.2 | 30/30 · 64.8 | 30/30 · 58.2 | 30/30 · 75.1 | 30/30 · 67.2 | 30/30 · 71.5 |
Supported-language review and examples needing caution
Conditions and limitations
- Results from November 14, 2026, using the same models and settings as the app.
- Sources and references were AI-authored, without native-speaker review. Wording different from the reference may still be correct.
- chrF measures character-sequence similarity to the reference (0–100; higher is closer). Correct paraphrases and tone differences can lower it. Do not compare scores across languages.
- “Output returned” only means a non-empty response, not acceptable quality. Pages also include untranslated English, wrong-language text, repetitions and truncated outputs.
- Only 30 short conversational sentences were tested. This does not establish performance on long text, names or recognition errors.
- Translation quality is not guaranteed. Do not rely on it for important conversations.