alphacep / alphacep/vosk-api

i want hinglish voosk model for my project

Offen
#1,860 0 Kommentare 2 Reaktionen 0 zugewiesene Personen Auf GitHub ansehen
Vorherrschende Sprache
Jupyter Notebook
Sterne
15.1k
Forks
1.8k
PR-Merge-Kennzahlen
Keine gemergten PRs in 30 T.

Beschreibung

To support precise offline recognition of Hinglish—informal speech blending Hindi grammar with English vocabulary—we employ a customized Vosk speech‑recognition model. This model is built by fine‑tuning a multilingual Vosk base on curated Hinglish datasets, containing typical code‑mixed phrases like “Mujhe weather check karna hai” or “Set an alarm for 7 baje”. By training on acoustic samples from Indian speakers and adjusting the language model to accept both Hindi and English words in context, the model achieves high accuracy without needing an internet connection—ideal for privacy-sensitive or low-connectivity environments. The result is a smooth, natural user experience: users can interact naturally in Hinglish and the system will transcribe intents reliably, enabling bilingual voice apps, assistants, or transcription tools tailored for the Indian market.

Beitragsleitfaden

Für dieses Repository ist kein Beitragsleitfaden indexiert

Bewertung

Dieses Issue wurde noch nicht bewertet.

Neue Issues direkt in Ihr Postfach

Eine kurze Übersicht über anfängerfreundliche GitHub-Issues.