Oyma's on-device Whisper engine supports over 99 languages out of the box, including Spanish, French, German, Japanese, Mandarin, and Arabic, with automated language identification and English translation.
Global Multilingual Speech Recognition
Modern teams and international research collaborations routinely conduct meetings across multiple languages or switch fluidly between languages during technical discussions.
Oyma leverages Whisper’s multilingual pre-training, which spans over 680,000 hours of multilingual and multitask supervised audio data.
Top Supported Languages
Whisper provides robust transcription accuracy across more than 99 world languages:
- European Languages: English, Spanish, French, German, Italian, Portuguese, Dutch, Swedish, Polish, Russian, Ukrainian.
- Asian Languages: Mandarin Chinese, Cantonese, Japanese, Korean, Vietnamese, Hindi, Indonesian, Thai.
- Middle Eastern Languages: Arabic, Hebrew, Persian (Farsi), Turkish.
| Language | Typical Word Error Rate (Large-v3) | Auto-Detection Support |
|---|---|---|
| English | < 4.2% | Instantaneous (< 1 sec) |
| Spanish | < 4.8% | Instantaneous (< 1 sec) |
| German | < 5.6% | Instantaneous (< 1 sec) |
| Japanese | < 6.1% | Instantaneous (< 1 sec) |
| Mandarin | < 6.8% | Instantaneous (< 1 sec) |
Key Multilingual Capabilities
- Automatic Language Detection: Whisper identifies the spoken language within the first 30 seconds of audio without requiring manual configuration.
- Speech Translation to English: In Preferences > Transcription, you can enable “Translate to English.” Whisper will transcribe spoken Spanish, German, or Japanese directly into fluent English prose in real time.
- Mixed Code-Switching: Handles bilingual speakers who alternate naturally between English and their native language during technical conversations.
Multilingual Accuracy and Dialect Handling
Whisper’s training on vast global speech corpora provides distinct advantages:
- Accent Robustness: Handles regional English accents (British, Scottish, Australian, Indian, Nigerian) with high word accuracy.
- Automatic Language Switching: When multilingual speakers transition between languages during technical reviews, Whisper detects and transcribes each language accurately.
- Specialized Vocabulary Support: In Preferences > Transcription, you can provide custom prompt words (e.g., industry acronyms, company names) to guide Whisper’s spelling of obscure domain terms.
Multilingual Translation Capabilities and Vocabulary Guidance
In addition to transcribing in the original spoken language, Whisper can automatically translate non-English audio directly into fluent English prose in real time, making international multi-language meetings effortless to review.
You can also pass custom dictionary prompts into Whisper to teach it specific acronyms, corporate product names, or medical terminology, ensuring uncommon words are spelled correctly throughout your transcripts.
Language Support Summary and Dialect Adaptation
- Coverage: 99+ world languages including Spanish, French, German, Japanese, Mandarin, and Arabic.
- Auto-Detection: Automatically determines spoken language within the first 30 seconds of speech.
- Live Translation: Optionally translates spoken foreign languages directly into English prose in real time.
- Domain Prompts: Custom prompt vocabulary ensures specialized corporate acronyms and terminology are spelled correctly.
- Code-Switching: Fluidly handles bilingual conversations where technical speakers alternate naturally between English and their native languages.
- Punctuation and Capitalization: Delivers complete grammatical capitalization, sentence boundaries, and comma placement across all supported languages.
