Fact · Languages and accents
29languages
The hayamimi README describes a language catalog of five routes (Japanese, Chinese, Korean, Cantonese and English) plus 24 EU languages, each sent to a dedicated model: 29 languages in all.
Quote found word for word on the live source, last checked October 21, 2026
In the source's own words
“5-route language catalog ja/zh/ko/yue/en+24 EU languages each go to a dedicated best-in-class model; everything else (~1600 languages) falls back to Meta's Omnilingual ASR”
hayamimi (早耳) README, oboroge0 on GitHub, oboroge0, GitHub, 2026, README, Features table.
Open the original- What it measures
- hayamimi is an open-source real-time speech-to-text tool. Its README says it runs on CPU only, with no GPU and no cloud API. Languages outside the catalog are sent to a separate model, Meta's Omnilingual ASR.
- What it does not show
- The 29 counts only the languages with a dedicated model. The README says other languages, about 1600, fall back to Meta's Omnilingual ASR, a separate model that this row does not count. The README's own comparison document holds the per-language results.
In the data
A row of How Many Languages Speech Recognition Supports, as hayamimi: 29 languages.
- System
- hayamimi
- Maker
- oboroge0
- Kind
- Dictation app
- What is counted
- Languages
- On device
- Yes
- As of
- 2026-10
- Note
- The quote gives 5 routes (ja, zh, ko, yue, en) and 24 EU languages; 5 + 24 = 29 languages.
Source
Used to calculate
- In the languages supported by speech recognition systems dataset, the 120 published figures for “Dictation app” have a median of 50 languages, ranging from 1 to 1,600.
- The languages supported by speech recognition systems dataset holds 266 data points, each a published figure traced to its source.
- The languages supported by speech recognition systems dataset draws on 221 distinct sources.