Where interview recordings come from
Most interviews end up as one of a few files. A phone voice memo app saves an M4A or MP3. A handheld recorder saves WAV or MP3 to an SD card. A video call leaves an MP4, and Zoom usually adds a separate M4A with only the audio. A video interview filmed on a camera gives you an MP4 or MOV.
You can upload any of these as they are. Audio is pulled out of video files on our side, so there is nothing to convert first. If you have both a video and an audio-only copy of the same call, upload the audio file. It is smaller, so it uploads faster, and the transcript is the same.
Recording tips that make the transcript better
The biggest gains come before you press record. Speech recognition copes well with a clear voice and struggles with distance, echo and background noise. A few habits help:
- Put the phone or recorder between the two of you, closer to the person you will quote most. A table mic at arm's length beats a phone left on a shelf.
- Pick a quiet room with soft surfaces. Cafés are hard because of music, cups and nearby voices.
- On a call, ask the other person to use headphones. This stops their speakers from feeding back into your recording.
- If you record a remote call, record your own side locally as a backup in case the connection drops.
- Say the date, names and spellings of unusual names at the start. It gives you a reference when you check the transcript.
Checking quotes and labelling who said what
When the transcript is ready, the player is synced with the text. The current line is highlighted as it plays and you can click any line to jump to that moment. This is the fastest way to confirm a quote word for word before you publish it or put it in a hiring report.
Fix names, company names and jargon by editing the text in place. Speech recognition has no way to know how your interviewee spells their surname, so check proper nouns every time.
Speaker labels are available on paid plans (Plus, Unlimited) or with a Minute Pack. They are not part of the Free plan or the no-sign-up trial. With speaker identification on, you can tell it how many people are in the recording, which helps when you already know the number. Then rename Speaker 1 to "Interviewer" and Speaker 2 to the guest's name once, and every line updates. Labels are most reliable when people take turns. If you talk over each other a lot, expect a few lines to need a manual check.
Which export fits your interview
Pick the export that matches what happens next. Timestamps and speaker names can be switched on or off in each export.
- Journalists: TXT or the Copy button to paste quotes into your draft. Keep timestamps on in your working copy so you can find each quote in the audio again.
- Hiring and HR: DOCX or PDF to attach to a candidate record or share with the hiring panel. A read-only share link works when colleagues only need to read it.
- Content teams turning an interview into an article or Q&A: DOCX with speaker names on and timestamps off reads like a clean script.
- Summaries: the AI prompt export copies the transcript together with a prompt you can paste into ChatGPT or Claude to draft a summary or pull out key points. RocketScribe itself does not write the summary.
Honest limits
A clear one-to-one interview usually transcribes very well. Accuracy drops with background noise, a phone left far away, crosstalk, strong accents and poor call connections. Always listen back to anything you quote directly.
Interviews in other languages work too. More than 90 languages are supported, auto-detected or chosen per file, and mixed Chinese and English speech is handled. Free accounts have a per-file length cap, so a long interview may need a paid plan. Your files are never used to train AI models, and you can delete them whenever you like.