Go to Apify Store
User picture

Luboš Benetka

soboljem

I build small tools that do one job properly. Currently: transcription that works on videos nobody else can caption.

ACTOR STATS

1 public Actor

2 total users

1 monthly user

What I build

Small, boring tools that do one job and do not lie about it.

Why my actors return warnings instead of confident nonsense. Speech-recognition models trained on subtitle files invent subtitle furniture when there is nothing to hear — translator credits, "thanks for watching", channel names. My transcriber strips those and flags the result as low confidence. An honest "we could not hear speech in this" is more useful than a fluent invention, and you can filter on it.

What I care about

  • Measured, not assumed. Every number in my listings comes from a run I can show you.
  • Failing loudly beats returning something plausible. A transcript with a silent gap in the middle reads perfectly and you have no way to tell — so my actors error instead.
  • Using the free path first. If a video already has captions, you should not be paying for speech recognition.

Found a bug or an edge case that breaks something? Open an issue on the actor — I read them.

Public Actors