Transcribe a video to text (Hebrew and about 100 languages)
Transcribe a video to text: upload it to StreamLive Pro, pick the same language as source and target, correct the lines and download TXT, or SRT with times.
Updated · 3 min read
To turn a video into text, upload it in Translate and subtitles and choose the spoken language as both the source and the target. The same language on both sides means a transcript, not a translation. Correct the lines beside the video, then download TXT — Plain text for plain text, or SRT — Subtitles if you need the times too. Transcription works in about 100 spoken languages, Hebrew included.
Before you start
A file with sound. The upload card names the accepted files (MP4, MOV, AVI) and the size limit. A video with no audio track is refused, because there is nothing to transcribe.
Minutes. Transcription uses the plan's translation minutes, equal to the video's length; amounts per plan are on the /pricing?lang=en. The trial includes 10 minutes, for videos up to 10 minutes long.
Audio decides the quality. A microphone close to the speaker gives far better text than a room microphone. Music under speech and people talking over each other lower it most.
Upload. Open Translate and subtitles and press Upload new video for translation, or drag the file onto the page. A video already in your account: Translate existing video.
Same language twice. Choose the spoken language under Source language and the same one under Target language. Press Start translation.
Wait for processing. The card shows Processing with a percentage and then Ready. You can leave the page meanwhile.
Read and correct. Open the project and go to Subtitles. Every line sits in a list beside the video; clicking a line plays that moment. Fix names and terms first with find and replace, then read the rest.
Mark speakers if there are several. Each line can carry a Speaker. Use Change Speaker so an interview reads as a conversation.
Remove filler words if you want clean text. In the properties panel, Smart Tools has AI Cleanup, which removes "uh", "um" and similar words. Skip it if you need a verbatim record.
Download. Press Export. In the Subtitles tab, TXT gives text without times; SRT or VTT keeps a time on every line; on paid plans, JSON keeps a time for every word.
The Translate and subtitles screen (test account).
Choose the language yourself.Auto-detect exists, but when you know the language, picking it avoids a wrong guess on a short clip or one that opens with music.
Upload the original file. A copy recorded from a speaker or re-compressed several times loses the detail the recogniser needs.
One language per video. A lecture that switches between Hebrew and English transcribes best in the language spoken most; check the passages in the other language by hand.
Rate the result. The workspace asks whether you are satisfied with the transcription. Your answer goes to us and helps us find weak spots.
What can go wrong
"Video too long" message. Each plan has a maximum video length for transcription, and the trial's is 10 minutes. Cut the video into parts, or see the /pricing?lang=en.