Skip to content

Add captions to a video

Recognise speech automatically, correct the captions and add them to your video.

Processed by our own AI server. Your text is processed using an AI model hosted on VersoFile's own server. It is not sent to third-party AI providers or retained as a saved document.

or drop it hereSupported formats: MP4, M4V, MOV, MKV, WEBM, AVI, 3GP

The video is uploaded over an encrypted connection and processed on our own server with the open-source Whisper model; it is deleted after processing and is never sent to a third party.

How to use Add Captions to Video

  1. 1

    Upload a video (up to 15 minutes) and choose the spoken language or let it be detected.

  2. 2

    Check the captions: correct the text, adjust timing, add or remove lines while previewing them on the video.

  3. 3

    Download SRT or VTT, or create a video with burned-in captions or a subtitle track.

About Add Captions to Video

Speech is recognised on VersoFile's own server with whisper.cpp and OpenAI's open Whisper model (MIT licence); the video never goes to a third-party service and is deleted after processing. A voice-activity detector runs first, so silence, music and noise are skipped instead of being turned into invented words, a known weakness of Whisper. Recognition supports Arabic, English, Turkish, Russian, Spanish and German and is good but not perfect: names, background noise, accents and overlapping speakers cause mistakes, so the editor lets you correct every line before using it. Burned-in captions are drawn with proper Arabic shaping and right-to-left order, kept inside the frame, and the video is re-encoded; a subtitle track keeps picture and sound untouched. Processing runs on a CPU and takes roughly as long as the video for recognition.

Frequently asked questions

How accurate are the captions?

In our tests with recorded read speech about 9 in 10 Arabic words and 19 in 20 English words were correct. Noisy audio or several speakers lower accuracy, so always review the captions.

What is the difference between burned-in captions and a subtitle track?

Burned-in captions are part of the picture and always visible. A subtitle track can be switched on or off by the viewer and does not re-encode the video.

Does it invent text when nobody is speaking?

We detect speech before recognising it, so silent or musical parts produce no captions. If no speech is found at all, you are told and can add captions by hand.

Is my video stored?

No. It is processed on our server and deleted after the result is downloaded or automatically within about 15 minutes. It is never shared or used for training.