mcut SDK

mcut SDK / transcription-local/src / AudioChunk

Interface: AudioChunk

Defined in: packages/transcription-local/src/chunking.ts:11

Chunked-streaming plan for long audio: fixed windows with overlap, merged back together on word timestamps. Whisper's native receptive field is 30s; processing window-by-window keeps the tab responsive (progress per chunk) and lets the worker skip non-speech windows and retry hallucinating ones without redoing the whole file.

Properties

endS

endS: number;

Defined in: packages/transcription-local/src/chunking.ts:15

Window end (exclusive), seconds.

startS

startS: number;

Defined in: packages/transcription-local/src/chunking.ts:13

Window start in the source audio, seconds.