VoxSpica Alexandr Makeenak

Records speech and transcribes audio files to text on your own computer. No account, no subscription, no ads, and the audio never leaves the machine: after the recognition model is downloaded, the program works with the network unplugged. The engine is VOSK (Kaldi ASR) running on the CPU - a GPU is not used and is not required. 33 recognition languages and 9 interface languages. Text appears while you speak, with pause and stop. The history of recognitions is kept in SQLite and is searchable, and punctuation and capital letters are applied. The portable build is a single executable in a zip: unpack it anywhere and run it. Python is not required and nothing is installed. On the first launch the program asks for the interface language and downloads a recognition model - about 50 MB for a small one, up to 1.8 GB for a large one. Per-language installers are published as well (English, Russian, German, French, Spanish, Italian, Chinese); each carries its own model, so it works offline straight after installing. The program is not code-signed, so Windows may warn about an unknown publisher.