← Esportscribe

How to caption esports VODs and co-streams for YouTube

Speech models are already good at esports commentary — research on League Worlds broadcasts puts Whisper near human accuracy overall. Almost every remaining error is a name: Kai’Sa becomes “kai sa”, ZywOo becomes “zy woo”, Dust II becomes “dust two”. Fixing names is exactly what a per-game pack does.

1. Export the audio

Drop the mp4 straight in, or export an m4a/mp3 for a faster upload. Anything under 45 minutes runs in the browser; split a long series by map.

2. Let it detect the game

After each run Esportscribe counts name hits per pack and applies the best match. If a cast covers two games, switch packs or use all five at once; nothing is re-transcribed.

3. Check the corrections panel

Every name, scoreline and acronym change is listed with a timestamp. Player handles with stylised casing (s1mple, m0NESY, cNed) are kept exactly; ordinary-word names (Train, Cache, Tracer) only change when the game context is nearby, so a transcript about a train station stays about a train station.

4. Download .srt and upload

YouTube Studio → Subtitles → Add → Upload file (with timing). For Twitch VODs, export to YouTube first or run the downloaded VOD through the studio. Caption text is indexed for search, so a correctly spelled “Nepomniachtchi”-class name like Praggnanandhaa or K’Sante is how people find the clip.

Caption a VOD