Subtitle Generator

Subtitles in three steps

1. Choose a video, or an audio file, and pick its spoken language, or let StemVocal detect it.
2. Play it back with the captions over the picture. Click any word to hear that moment again, and correct names, numbers and anything misheard.
3. Download Subtitles (.srt), already selected, or switch to Web captions (.vtt).

Free subtitles cover recordings up to 30 minutes. For a longer film, lecture or livestream recording, credits handle files up to 3 hours and can add speaker labels.

Timed for reading

Subtitles should be readable at the speed people speak. StemVocal times each cue from the words' own timestamps rather than from fixed blocks. A cue stays on screen for at most 6 seconds and holds at most 80 characters, and anything longer than 42 characters splits into two balanced lines. Cues never overlap, so a player never stacks two captions at once.

Your edits come with you. When you rewrite a sentence, its words are spread across the time the original speech took.

SRT or VTT?

Choose SRT for a video editor or a site's subtitle upload form. Choose VTT for a web player that reads WebVTT captions. Both files hold the same cues and text; neither changes the video itself. For a written transcript instead, open video to text and download Word or plain text.

Subtitles in another language

A paid transcript can be translated with the export menu still on SRT or VTT. The translated file keeps every cue's timing, so it lines up with the video like the original. Translation runs from English into 15 languages, or from any of them into English. It is machine translation: have a speaker of the language review it before you publish.

When captions drift

Most timing problems come from changing the video after the subtitles were made. Generate subtitles from the final cut; if you trim the start or remove a section later, run the edited video again rather than shifting the old file by hand. If a line breaks a phrase awkwardly, edit the words in the transcript: the next download times the cue again from the same speech.

Check before you publish

Watch the whole video with captions on, full screen included, before you upload the file. Pay attention to names, numbers and places where people talk over each other, where the transcript usually keeps one voice. Every correction you make in the editor is in the next download.

1
00:00:00,000 --> 00:00:04,000
What should we check before publishing?

2
00:00:04,000 --> 00:00:10,000
Check names and numbers
against the recording. Then save

3
00:00:10,000 --> 00:00:12,000
the corrected transcript.