Vosk

• Vosk is an offline automatic speech-recognition (ASR) toolkit based on Kaldi and Vosk models.

• It is designed for low-latency streaming recognition and can run on relatively lightweight hardware such as Raspberry Pi, Android and other embedded systems.

• It supports 20+ languages and dialects, including English, Indian English, Hindi, Gujarati, Telugu, German, French, Spanish, Chinese, Russian and Japanese.

• Models range from small models of roughly 40–50 MB to much larger server models.

• It supports configurable vocabularies and speaker identification, making it useful for specialized voice applications.

• Country availability: No specific country restriction is publicly stated. Because Vosk runs locally, it is generally available internationally wherever the software/models can be downloaded.