This process ensures that subtitles are precisely aligned:
- Format Selection: Prioritizes the use of uncompressed WAV format (16bit/44.1kHz) to reduce decoding latency.
- parameter calibration: Add the following to the /srt interface
?word_timestamps=trueGet verbatim timestamp - manual calibration: Load audio waveforms with tools like Subtitle Edit to fine-tune keyframes.
- fault tolerance: When a worker times out and returns incomplete data, use the
start_time=上次结束时间Continuation of the remainder of the transmission
Final deviation can be controlled within ±200ms.
This answer comes from the articleWhisper on Cloudflare AI: a free tool to convert audio to text and generate subtitlesThe
May not be reproduced without permission:AI productivity tools " A guide to completely solve the problem of unsynchronized subtitle timeline
English 
简体中文 
日本語 
Deutsch 
Português do Brasil