Why this tool is useful
Turn speech in an owned audio or video file into editable text with a pinned multilingual model that runs only in this browser.
Local File Tools · runs locally
Private Audio to Text Transcriber
Turn speech in an owned audio or video file into editable text with a pinned multilingual model that runs only in this browser.
Ready. Your working data stays in this browser.
Files and capture streams stay on this device. Processing uses browser APIs only; nothing is uploaded to YTSave, the media API or a proxy. No server copy is created, stored or left to delete. The file picker is cleared when processing starts, and disposable worker memory is destroyed after completion or cancellation. Preview and generated-output memory is released when you clear the tool or leave the page.
How it works
The browser decodes and mixes the selected recording to mono 16 kHz audio, then a locally hosted quantized Whisper model performs bounded speech recognition in WebAssembly.
Practical example
Choose a short interview clip, optionally select its spoken language, review the transcript for recognition mistakes, and download plain UTF-8 text.
Validation
The calculation is deterministic and explains validation errors instead of silently changing invalid input.
Privacy boundary
Input and output remain in this tab. Copy and download happen through browser APIs without a server upload.
Result passport
Every run reports its tool mode, timestamp and input/output size so transformed data can be audited.