Batch convert videos to SRT offline

Example: a folder of course videos

Suppose you have three course videos, including one long recording. Start with a short file to confirm the selected Whisper model recognizes the source language, then queue the whole folder. This is a walkthrough example.

1. In Settings → Speech transcription, choose an engine and Whisper model and download the weights. Select Transcription only.
2. Return to the workspace and use Add folder to include files in subfolders. If the folder already contains subtitles, check the setting to skip videos with existing subtitles before interpreting the queue.
3. Start the batch and watch the queue. When a job finishes, find the SRT next to the source media. Check the beginning, end and a few important timecodes.

Resume an interrupted long video

Reopen the app and review the restored queue. Open the job’s More menu, choose Resume processing (keep progress), and start the batch again. The app reuses compatible checkpoints. If the file, source language or primary model changed, follow the resume or restart prompt. Restart clears the checkpoints for that job.

See what an exported SRT looks like

This four-cue source-language example shows numbering, timecodes and subtitle text. It uses the website’s demonstration script.

1
00:00:00,000 --> 00:00:04,000
Early in the morning, we take a slow walk down the street.

Set up a model for polishing or translation

To polish or translate subtitles, install Ollama and download a model, or supply your own API key for an OpenAI-compatible cloud service.