Video Transcriber
Transcribe a video or audio file with OpenAI's Whisper, running entirely in your browser. Your file is never uploaded β the only download is the Whisper model itself, which your browser keeps for next time.
Bigger models are more accurate but slower, and are a bigger one-off download. Your browser keeps downloaded models, so later transcriptions start straight away.
Turbo is the recommended balance of speed and accuracy. Turbo (best quality) is less compressed, for a little more accuracy.
Without a graphics chip your browser can use (WebGPU), everything runs on the processor and is much slower.
Auto-detect listens to the first 30 seconds to work out the language.
Choosing the language yourself helps with short clips, strong accents, or a video that starts with music or silence.
Text is the whole transcript as one block. Timestamped puts the start time in front of each passage.
SRT and VTT are caption files for video players and editors, split to the usual subtitle size: up to two lines of 42 characters.