Turn a video file into subtitles in another language

Why a long video trips up a chat assistant

A chat assistant can often caption or translate a short clip well. A film, a sermon or a two-hour family video runs into limits: assistants that accept audio cap how long a file can be, free plans often cannot upload audio at all, and a complete subtitle file for a long video does not fit in one reply. What comes back tends to stop partway or turn into a summary. Timing is the other gap, because an assistant cannot reliably say when each sentence starts in a long file.

This tool cuts the audio into parts of up to ten minutes. In each part the AI listens and writes timed cues, the times are moved by where that part starts, and the parts are joined in order. Code then checks that the cues run in order and do not overlap. For a translation, every cue is translated separately and keeps its start and end time.

What people subtitle

• Family videos and weddings, with English subtitles for relatives who do not speak the language
• Church services, sermons and religious talks for a congregation that speaks another language
• Lectures, courses and webinars for students in other countries
• Interviews and documentaries going to a festival, a client or a wider audience
• Long recordings in Amharic, Burmese, Pashto or Tigrinya that need subtitles in a language more viewers read

Using the SRT

In VLC, open the video and choose Subtitle > Add subtitle file. Plex reads an SRT saved next to the video under the same name. In YouTube Studio, open Subtitles, add the language and upload the file with timing. Premiere and DaVinci Resolve import an SRT as a caption track. Use the VTT where a web player asks for one.

Bilingual output puts the spoken line above the translation inside the same cue, which suits language learners and mixed audiences, but it takes more room on screen. If you need both languages as separate tracks instead, a translated job already gives you the translated SRT and an SRT in the language spoken.

Accuracy and limits

Timings come from the AI listening to the audio. They are close to the speech but not frame-exact, so for broadcast or festival delivery open the SRT in a subtitle editor and adjust the cues that matter. Music under speech, crowd noise and people talking over each other cause mistakes; check names and numbers before you publish.

The subtitles and the translation are made by AI models. Treat them as a draft to review, not a certified or sworn translation. Each job takes one file, so start a separate job for each episode. More subtitle languages may be added; the menu in the upload box always shows the current list.