Questions and answers about Soz

What does Soz do?

Soz recognizes spoken Kyrgyz in video, creates editable subtitle lines, and exports MP4, SRT, ASS, or TXT.

Can I paste a YouTube link?

Yes, when the video is accessible and you have the right to process it. If the link fails, upload an available video file.

Which video files can I upload?

The current interface lists MP4, MOV, and MKV. The file must also meet the service limits.

Can I process only part of a video?

Yes. Set the start time and duration of the fragment you need.

Do I need to review the recognized text?

Yes. Recognition provides a working draft, while names, dialects, noise, music, and uncommon words may need correction.

Can I edit the subtitles?

Yes. Edit each line in the review screen and check it against the video.

Which output formats are available?

MP4, SRT, ASS, and TXT. Available actions may depend on the project state.

Which format should I choose?

Use MP4 for visible subtitles inside the video, SRT or ASS for a separate subtitle track, and TXT for plain text.

Can I share a result with a public link?

A share action is available after completion. Create a public link only when you intend to give access to anyone who has the URL.

Is the source video used to train the model?

Source video and audio are not used for training without separate consent. An optional consent may cover corrected text and related technical data.

Why can processing take longer?

Time depends on fragment length, source availability, selected quality, and queue state.

What should I do if recognition is inaccurate?

Open the editor, correct the lines, and compare them with the audio. For difficult material, select a cleaner speech segment.