Transcribe Any Audio or Video File to Searchable Text

LiveScript accepts the common audio and video formats: mp4, mov, mkv, webm, avi, m4v, flv, wmv, mp3, m4a, wav, aac, flac, ogg, oga, opus, and wma. We extract the audio automatically, so you can drop in a video file and get a transcript of the speech without converting it first.

Free accounts can upload files up to 200 MB each. Pro and Day Pass accounts can upload files up to 5 GB each, which covers multi-hour recordings and high-bitrate video.

Uploads draw from the same monthly minute pool as the live transcript API. The Day Pass includes about 1,000 minutes and Pro includes about 10,000 minutes per period, and top-up packs (500 / 1,500 / 5,000 minutes) stack on top. Free accounts get a starter pool of minutes to try it. You're only charged for the duration of audio actually transcribed.

Uploads are transcribed with the same speech-to-text engine that powers our real-time live transcripts, tuned for clean recorded audio. Clear single-speaker recordings are highly accurate; heavy background noise or overlapping speakers reduce accuracy, as with any transcription tool.

Uploads are diarized automatically: the engine clusters the audio by voice and tags each line with a speaker, which you see as Speaker 1, Speaker 2, and so on. It works from the sound of the voices alone - there is nothing to set up beforehand, no voice samples to provide, and no need to tell us how many people are in the recording. You then put real names on the speakers, and those names appear throughout the transcript and in your subtitle exports.

It is a best-effort automatic pass, not a guarantee. Clean recordings with distinct voices and little crosstalk separate well. Overlapping speech, similar-sounding voices, or several people sharing one microphone can cause lines to be attributed to the wrong person, or split one person across two labels. Because of that, every part of the transcript is editable: reassign a line to a different speaker, split a line where the speaker actually changes, add a speaker the automatic pass missed, or merge two labels by giving them the same name. Speaker numbering is per file, so the same person in two different uploads is not linked automatically.

Not currently. Speaker identification runs on uploaded audio and video files. Live stream transcripts and YouTube backfills are transcribed without speaker separation.

Yes. Multilingual mode handles Spanish, French, German, and many other languages, including recordings that mix languages. English-only mode gives the highest accuracy for English-only audio.

Yes. Open a completed upload to read the timestamped transcript, click any line to scrub the built-in audio player to that moment, and download subtitles as SRT or WebVTT for editors, captioning tools, or your own archive. You choose whether speaker names are written into the captions, so you can hand the same recording to a video editor with names in and to a subtitle workflow with them left out.

Your transcript stays in your account so you can search and revisit it. The uploaded audio file itself is automatically purged after a retention window (30 days by default) to save space - the transcript remains available after the audio is removed.

Uploads are private to your account by default. You can optionally generate a share link for an individual upload if you want to send the transcript and audio to someone else; otherwise no one else can access it.