A live translation is gone the moment it is spoken. That is fine while you are listening and a problem the next morning, when you need the sentence somebody actually said, the spelling of a name, or a subtitle file for the recording you made alongside the session. VoxisLive keeps a record of every session so those questions have an answer that does not depend on your memory.
This page describes what that record contains, what you can do with it, and what it deliberately does not do.
What a session record contains
While a session runs, VoxisLive accumulates the translated lines it produced and, where they are available, the source lines they came from. At the end of the session that accumulation is written to disk as a single record, together with the timing each line carries.
From that one record everything else is derived: the exports, the history entry you search later, and any corrections you make. There is no separate "save" step to remember and no second copy to keep in sync.
Exports: TXT, SRT and VTT, bilingual or translation-only
A record can be exported in three formats, and each of them in two variants: bilingual, which keeps the source line next to its translation, or translation-only, which keeps just the translated text. Each variant gets its own filename, so exporting both does not overwrite the first one.
| Format | What it is for |
|---|---|
| TXT | Plain reading and pasting — a meeting summary you write yourself, a quote for an email, a search in a text editor. |
| SRT | Subtitles for a video file, in the format most players and editors accept. |
| VTT | Subtitles for web video, for a player that expects WebVTT rather than SRT. |
Exports land inside the session's own folder, beside the record they came from, so a session you archive or send on stays complete.
Subtitle lines fold at word boundaries, and are never truncated
Subtitle formats expect short lines. A naive exporter hits its character limit and cuts — and the words after the cut are simply gone from the file, which you discover while watching the video back. VoxisLive folds instead: a long line is broken at a word boundary and continues on the next one.
The practical guarantee is narrow and worth stating plainly: nothing is dropped in the fold. Every word that was in the record is in the exported subtitle file, wrapped rather than shortened.
The timeline is anchored to the start of the session
Cue offsets in the SRT and VTT files count from the moment you pressed start, not from the first word anybody happened to say. If you also recorded the session — with the app's own dual-track recording, or with OBS, or with the conferencing tool's recorder started at the same time — the subtitle timeline and the recording share an origin, and the file drops onto the video without you hunting for an offset.
Simultaneous translation runs behind the speaker by its nature: a translated line is finished, and therefore timestamped, a few seconds after the source words were spoken. The anchor makes the timeline line up with your recording; it does not turn a simultaneous interpretation into professionally timed subtitles.
History: searchable, one folder per session
Saved sessions appear in the app's history, newest first. Typing in the search box filters the list by what was actually said, and the matching text is highlighted where it occurs, so you can tell why a session matched before you open it.
Each session lives in its own folder. The record, any exports you generated and any audio you recorded sit together. That layout is what makes the two simple operations behave the way you would expect:
- Export writes the new file into that session's folder, next to the record.
- Delete removes the session's folder as a whole, so a deleted session does not leave exports or audio behind on disk.
Star a session and housekeeping leaves it alone
Transcripts accumulate. VoxisLive does routine housekeeping on old ones so the folder does not grow without limit. Starring a session is the switch that opts it out: a starred session is exempt from automatic housekeeping.
That is the whole feature. It is not a folder system, not tagging, and not a favourites view with its own organisation — one flag, one consequence, on the sessions you cannot afford to lose to a cleanup you were not watching.
Correcting a line: inline editing with narrow write access
Speech recognition and machine translation both get names, jargon and numbers wrong sometimes. You can edit a saved record in the app rather than exporting it and fixing the same mistake in three files.
What is writable is deliberately limited to two things: the translated text and the source line. Timing and speaker fields are not editable and are not rewritten when you save a correction — the structure of the record stays exactly as it was recorded, and only the words change.
Because every export is generated from the saved record, a correction you make once flows into every file you export afterwards. There is no re-export ritual to remember and no risk of a corrected TXT disagreeing with an uncorrected SRT.
Auto-save every two minutes while the session is live
You do not have to finish a session cleanly to keep what it produced. While a session is live, the record is written to disk every two minutes. A crash, a power cut, or a laptop that decides to reboot for updates during a three-hour call costs you the last stretch, not the whole session.
The periodic write covers the record itself. Exports are produced when you ask for them, from whatever the record holds at that point.
Recording the audio: two tracks, not a mix
Alongside the text record, VoxisLive can write the audio as two separate WAV files — the source audio in one, the translated speech in the other. Two files rather than one mix, because a mix cannot be undone: with separate tracks you can check a disputed sentence against what was actually said, edit either side independently, or discard one and keep the other.
Two limits on this one, both intentional:
- It is optional and off unless you turn it on.
- It is available in Video mode only. Recording a meeting is a different act from recording a video you are watching: in most places it needs the consent of everyone on the call, and that is a decision for you and the people in the room, not a checkbox we hand you inside Meeting mode.
The files are on your disk
Transcripts are stored on your own machine. Searching your history reads local files; exporting writes local files; deleting a session removes a local folder. There is no transcript library in an account that you have to trust us to keep, and nothing to migrate if you stop being a customer.
If you would rather verify that than take our word for it, the capture and local storage code is published as a read-only excerpt — see licensing for what that repository is and is not.
What this is not
- It is not a subtitle timing tool. The cues follow a simultaneous interpretation, anchored to the start of the session; they are not frame-accurate captions.
- It is not a cloud transcript archive with sharing links, comments or team access. The record is a set of files in a folder on your computer.
- Starring is not a folder or tagging system — it only exempts a session from automatic housekeeping.
- Editing is not a full record editor: text and source lines are writable, timing and speaker fields are not.
- Dual-track recording is not available in Meeting mode, by design.
For the rest of what the app does during the session itself, see features; for what a session costs, see pricing.