transcrito.app

The application runs on Windows and lets users convert audio or video files into text without sending any data to the cloud. It uses OpenAI’s Whisper model through the Whisper.net wrapper, extracting the audio with FFmpeg and performing inference locally, either on the CPU or, if available, on a GPU via Vulkan. Users select a file, optionally lock a language or let the model auto‑detect it, choose a Whisper model size, and start transcription; the result is saved as a plain‑text transcript, a subtitle file, or both in the same folder.

It supports virtually all common media containers—MP3, MP4, WAV, FLAC, MOV, M4A, WebM, MKV, Opus, Ogg, etc.—thanks to FFmpeg, and can handle 99 languages with the underlying Whisper model. Model files are downloaded once from Hugging Face and cached, and the app reports only a single anonymous ping per installation for usage statistics.

The software is open source, built with .NET 8, and emphasizes privacy and offline operation: no uploads, minimal telemetry, and the ability to run entirely on a local machine, with optional GPU acceleration for faster processing.