Introducing TehTranScriber

Today we're introducing TehTranScriber, a new uploader on The Subtitle DB. It creates subtitles from a movie's or episode's own audio using Whisper, the open-source speech recognition model from OpenAI. Every file it uploads is marked "AI transcribed," so you'll always know where a subtitle came from.

How we decide what to transcribe

Once a day, we rank movies and episodes by popularity. We look at IMDb votes, TMDB popularity, and how many people downloaded subtitles for the title here over the last 30 days. Then we check what we already have. A title is picked if it has fewer than three subtitles, if it has none in its original language, or if it has none in a language its audience is asking for.

To find out which languages people want, we count each title's audience by country over the last 30 days and match each country to a language. A language needs at least 50 people and 5% of the title's audience before we act on it. Half of each day's list is set aside for these language requests, so they don't get stuck behind everything else.

What it can and can't do

TehTranScriber writes down what is actually said, so it doesn't translate. To make Portuguese subtitles, it needs a Portuguese dub of the title. If there isn't one, the title waits until there is. Before transcribing an episode, it checks that the file really is that episode, and it skips any audio track it isn't confident about.

The first 59 transcripts went live on October 23, 2026, and new ones are published within minutes of being transcribed.