Upload the MP4 as it is, or import it from a link
You do not need to convert an MP4 to MP3 before transcribing it. The audio is pulled out of the video on our side, and only the audio is sent to speech recognition. The video picture is not used.
The video track is most of an MP4's size, so a long recording can be a large upload. Uploads are sent in parts, and if the connection drops, Retry continues where it stopped instead of starting over. If a file is above your plan's size limit, exporting just the audio from your editor or player gives a much smaller file with the same transcript.
Meeting recordings often already live in the cloud. Instead of downloading a large MP4 and uploading it again, paste a Google Drive or Dropbox link into the import box. The link must allow access to the file, for example a link set to anyone with the link. Direct links to MP4 files on your own server work too.
Common sources of MP4 files
MP4 is the default video format for most meeting tools, phones and screen recorders. Typical files people transcribe include:
- Zoom recordings, both cloud and local.
- Microsoft Teams meeting recordings.
- Google Meet recordings, which are saved to Google Drive as MP4.
- Videos recorded on Android phones.
- Screen recordings and tutorials from tools like Loom or OBS.
- Webinars and course videos you have downloaded.
Videos with more than one audio track
Some MP4 files carry several audio tracks. A screen recorder may save your microphone and the computer's sound separately, and some meeting tools record each participant on their own track. When a file has more than one audio track, all tracks are mixed together before transcription, so nobody's voice is dropped.
The one case where this works against you is a video with alternate language tracks, such as an English and a Spanish dub of the same film. Mixing those would overlap two languages. Remove the extra tracks in your editor first and keep only the one you want transcribed.
Plain text or subtitles: choosing the output
For meeting notes, interviews and research, export the MP4 transcript as Word, PDF or TXT. Timestamps and speaker names can be switched on or off in the export. Speaker labels are available on paid plans. For a long meeting, the AI prompt export copies the transcript together with a prompt you can paste into ChatGPT or Claude to get a summary or a list of action items.
If you want captions, export SRT or VTT. Edit any wrong words in the editor first, since the subtitle file is built from the corrected text. Upload the file to your video platform or import it into your editor. RocketScribe produces the subtitle file only. It does not burn captions into the video.
What affects accuracy in video
Video audio is often harder than a dedicated voice recording. Edited videos may have background music under the speech. Phone videos rely on the built-in microphone, which may be several meters from the person talking. Meeting recordings pass through video-call compression and include crosstalk when people interrupt each other. Clear speech near the microphone gives very good results, and these conditions reduce accuracy. Plan to check names, numbers and technical terms against the video before you publish anything.
Zoom can also save a separate audio file for each participant when that option is turned on. Each of those files is transcribed on its own. If you want one transcript of the whole meeting, upload the combined MP4 instead.
Speech in 90+ languages is supported. The language is detected automatically, or you can set it for each file. Meetings where people switch between Chinese and English are handled as well.