WhisperWebUI

I built WhisperWebUI to make Whisper transcription approachable for everyone: upload audio or video in the browser, get multilingual text, and export subtitles or notes without installing a desktop app. I am especially interested in feedback from creators, students, and teams using it for real workflows.2026-08-03

The thing I'd want to know as a user is how you handle audio longer than Whisper's 30 second window. Chunking is where most browser implementations get rough: boundaries land mid word and timestamps drift over an hour long file, so subtitles slowly desync. If you're doing overlapping windows and stitching on token overlap, that's worth saying on the page, because it's the difference between subtitles people can ship and ones they have to hand fix. Also curious whether you fall back to WASM when WebGPU isn't there, and what that does to the first run model download.

Does the transcription run fully in the browser or is the audio sent to a server? For meeting recordings that distinction matters for privacy, and it also decides how long a one hour file takes. Also, SRT export is the part I care about most: can users cap characters per caption line, or is segment length whatever Whisper returns? Default Whisper segments are usually too long to fit on a vertical video.