How to transcribe an interview on Windows without uploading it

Journalists, researchers, podcasters and HR teams all end up with the same job: a recording of a conversation that needs to become text you can quote. Online transcription services do it well, but they need the audio uploaded to their servers — not ideal when your interviewee spoke on condition of confidentiality, or your research ethics approval says recordings stay on your device.

This guide uses TemaMoji, a Windows transcription app from TemaLab that runs on your PC — your recordings are never uploaded. The free version transcribes the whole interview so you can check it on screen; exporting and copying cover the first 10 minutes.

1. Before you record

• Put the microphone between you and the interviewee, or use a separate mic for each person. Distance and background noise affect accuracy more than anything else.
• Try not to talk over each other. Overlapping speech is where speaker labels most often go wrong.
• Any common format is fine: a phone voice memo (m4a), a recorder (mp3, wav), or a video call recording (mp4).

2. Transcribe the recording

1. Install TemaMoji (Windows 10/11). The app opens in English if Windows is in English.
2. Under Transcription settings, set Spoken language (English, Japanese or Chinese) and Number of speakers. For a one-on-one interview, choose 2.
3. Drag the recording onto the window, or click Choose files. The first time, the transcription model (about 570 MB) is downloaded once.

Progress and the time left appear in the queue, and you can watch the text come in.

Without an NVIDIA GPU, transcription takes roughly as long as the recording. With an NVIDIA GPU, adding the GPU version from the app makes it several times faster.

3. Name the interviewer and interviewee

When it's done, the transcript opens split into paragraphs by speaker. In Speaker names on the left, rename “Speaker 1” and “Speaker 2” — for example to “Interviewer” and the interviewee's name. Every paragraph updates.

If one paragraph has the wrong speaker, click its speaker name and pick the right one. If one person was split into two labels, give both the same name and they merge.

4. Check quotes by ear

Before you quote anyone, listen. Click a paragraph and playback starts from that moment; edit the text as you listen. Ctrl+K plays and pauses, Ctrl+J and Ctrl+L skip back and forward 5 seconds, and you can slow playback down to 0.5×.

The Remove fillers switch hides “um” and “uh”. Leave it off when you need a verbatim transcript.

5. Fix names and terms everywhere

Names, products and jargon are the usual misses. Press Ctrl+F, type the wrong spelling and the right one, and click Replace all.

6. Export

Click Export and choose:

• Word (.docx) — laid out as a transcript, each speaker's name and time above what they said. Good for editing and sharing.
• Text (.txt) — for pasting into your notes or a writing app.
• Subtitles (.srt) — if the interview is a video you're publishing.
• CSV — start, end, speaker and text columns, for analysis in a spreadsheet.

Save the TemaMoji file too, so you can reopen the transcript and keep editing later.

7. Limitations

• Accuracy depends on the recording. Phone calls, echoey rooms and heavy accents produce more mistakes, so always check quotes by ear.
• Similar voices and crosstalk can confuse the speaker labels.
• TemaMoji works on recordings. It doesn't transcribe live or join calls.
• Windows only.

Published: October 27, 2026 · TemaLab