ক্যাপশনBangla subtitles

Checking the server

Bangla subtitles from your video

Upload a video or audio file. Our open Bangla speech recognition model writes an .srt subtitle file with timestamps.

How it works

  1. Voice detection finds the speech and cuts it into pieces of up to 25 seconds.
  2. Each piece is transcribed by our Whisper-medium model, fine-tuned on about 300 hours of Bangla speech.
  3. On whole media recordings our fine-tuned models get about 14% of words wrong, against 26% for the model they started from.
  4. Your video is deleted from the server as soon as the subtitles are made.