How to transcribe an interview (fast, free, accurate)
A practical guide to transcribing interviews in minutes, upload audio or video to AI, convert Zoom or Teams recordings, handle phone interviews, and export clean transcripts.
By Serhii MaysonPublished Updated 7 min read

Manual transcription of a one-hour interview takes between four and six hours. AI transcription is far faster. In October 2026, half of the uploaded files were transcribed in 14 seconds or less, and a 145-minute recording took 2 minutes 45 seconds. Times don't include uploading, which depends on your connection. More on long recordings of 1–3 hours. That gap is why almost every working journalist, researcher, and content creator has switched, the question is just how to set it up correctly the first time.
Transcribe your interview recording here
How do I transcribe a 2-hour interview recording?
Upload the recording (MP3, M4A, WAV or MP4, up to 2 GB). The first 30 minutes are free, and your first file needs no account. Transcribing all 2 hours costs $2 once. That covers one recording up to 3 hours. For several interviews, a $5 Week Pass gives 7 days of Pro and never renews. In October 2026, a 145-minute recording took 2 min 45 s once uploaded. TranscribTxt doesn't label speakers, so add names during review.
This guide covers three common interview scenarios, what makes transcription accurate or inaccurate, and how to get the most useful output from your transcript.
Why transcription accuracy varies (and what you control)
Before the method, it helps to understand the one factor that matters most: audio quality. AI transcription models convert speech to text based on acoustic signal. A clear recording with one speaker close to the microphone comes back with few errors. A phone call recorded through a laptop speaker in an echoing room comes back with many more.
You can't fix bad audio after the fact, but you can make better recording decisions before the interview. More on that in the tips section.
Method 1: Upload an audio or video file to TranscribTxt
This is the baseline workflow and works for any recording you already have on your computer.
Step 1: Go to TranscribTxt's interview transcription tool and upload your file.
Drag the file into the upload zone or click to browse. Your first file needs no account; a free account covers 3 files a week, up to 30 minutes each. Supported formats include MP3, M4A, WAV, MP4, MOV and WebM, up to 2 GB per file on every plan. An interview longer than 30 minutes gets its first 30 minutes free; the whole file (up to 3 hours) is $2 once; for several interviews, use a $5 Week Pass (7 days of Pro, no renewal) or Pro.
The file is uploaded over HTTPS.
Step 2: Let it detect the language.
There's no language menu: TranscribTxt detects the spoken language automatically across 99 languages, including English, Spanish, French, German, Portuguese, Russian, Arabic, Japanese, Chinese (Mandarin), Korean, Hindi, and many more. If the recording opens with a long silence, music or noise, trim it before uploading so detection hears speech first.
Step 3: Wait for processing.
Longer files take longer. You'll see upload progress, then a "Transcribing…" status.
Step 4: Review and export.
Once processing is complete, the transcript appears on screen. Use the Copy button to paste directly into a document, or click Download .txt to save the file.
Your uploaded file is deleted after processing (a free-preview upload is kept up to 24 hours so you can finish the rest). We don't keep transcripts on our servers either, so save the text before you leave the page.
Every plan can also download .srt, a subtitle file with timestamps, useful for captioning a video version of the interview. Pro ($10/month) adds JSON with word-level timestamps. TranscribTxt doesn't label speakers; see the review tip below for adding names quickly.
Method 2: Zoom or Teams recorded call
Remote interviews recorded through Zoom, Microsoft Teams, or Google Meet produce a video file you upload exactly like any other.
Getting the file from Zoom:
Local recordings save automatically to Documents > Zoom > [Meeting name] on your computer. The file is named zoom_0.mp4. If you recorded to the Zoom cloud instead, log in at zoom.us, go to Recordings, and download the MP4.
Getting the file from Microsoft Teams:
Teams saves cloud recordings to SharePoint or OneDrive (check with your IT team for the exact location). From the meeting chat, click the recording thumbnail and choose Download.
Getting the file from Google Meet:
Recordings save to the organizer's Google Drive under a folder called "Meet Recordings." Download the MP4 from there.
Once you have the file, upload it to TranscribTxt the same way as any other video. A 60-minute call is over the free plan's 30-minute limit: you get the first 30 minutes free, and the whole call is $2 once.
One note: recorded video calls often have audio quality issues that reduce accuracy, participants on weak internet connections, people speaking while muted, room echo from laptop speakers. Expect noticeably more errors than from a dedicated microphone recording.
Method 3: In-person interview recorded on a phone
Phone recordings are the most common interview scenario for journalists and researchers working in the field.
iPhone: The Voice Memos app saves recordings as M4A files. After the interview, open the Memos app, tap the recording, tap the three-dot menu, and select Share. AirDrop it to your Mac or email it to yourself.
Android: The Voice Recorder or Recorder app (varies by manufacturer) saves as MP3 or M4A. Use Files or Google Drive to transfer to your computer.
Once the file is on your computer, upload it to TranscribTxt. M4A and MP3 both work without any conversion.
Accuracy tip for phone recordings: The placement of the phone during recording makes the biggest difference. A phone lying flat on a table between two speakers picks up both voices reasonably well, but the person sitting farther from the phone will be quieter in the recording. If the imbalance is significant, the AI model may struggle with the quieter voice. A better setup: hold the phone close to each speaker as they talk, or use a small portable recorder with dual microphone inputs.
Tips for better transcription accuracy
Record one speaker at a time. Overlapping speech is where every AI model makes the most errors. Brief interruptions are fine; sustained crosstalk produces inaccurate output. If you are conducting the interview, train yourself to wait a beat before responding. The same habits help when you record a conversation between two or three people; here's how to transcribe a conversation once it's recorded.
Use a microphone, not a laptop speaker. Even an $30 USB microphone placed 30–40 cm from the speaker produces significantly cleaner audio than a laptop's built-in mic from across a desk.
Avoid background noise. Coffee shops, open offices, and outdoor locations introduce noise that reduces accuracy. A quiet room with soft furnishings, carpet, curtains, upholstered chairs, absorbs echo and produces noticeably cleaner recordings.
State names and technical terms clearly. AI models make the most errors on proper nouns: names, places, organization names, technical jargon. If your interview involves terms the model is unlikely to have encountered, enunciate them clearly and plan to review them manually.
Review at speed. Use VLC media player set to 0.75x speed and follow along with the transcript. Read the AI output and correct as you go. For a 45-minute interview, this review pass takes about 15 minutes. Focus your attention on proper nouns and anything that looks phonetically plausible but wrong.
Label speakers as you review. TranscribTxt doesn't separate speakers, so add names during your review pass. In a two-person interview, questions and answers usually alternate, which makes this quick; keep the SRT open to jump to any line you're unsure of.
Exporting: which format to use
Plain text (.txt) is the right choice for most interview workflows: notes, quotes for articles, qualitative coding, meeting summaries. It is clean, editable in any text editor or word processor, and easy to paste into Claude or ChatGPT for summarization. Available on the free plan.
SRT subtitles (.srt) add timestamps to every line of text in a format video editors understand. If your interview will be published as a video, on YouTube, as part of a documentary, or as a social media clip, SRT captions make the video accessible and improve retention. SRT export is included on every plan, free too. You upload the .srt file directly to YouTube Studio or drag it into your timeline in Premiere Pro or DaVinci Resolve.
JSON with word-level timestamps is the format to use if you are building anything programmatic, a searchable archive of interviews, a transcript review tool, or an integration with a research database. It includes every word with its start and end time, but no speaker identifiers. JSON export is on Pro ($10/month).
Common mistakes to avoid
Trusting the output without review. Even at 97% accuracy, you'd get roughly three errors per 100 words. In a 5,000-word transcript, that is 150 errors. Most are minor (a word substitution or punctuation issue), but a misheard name or number can create a meaningful factual error if you quote it directly.
Starting the recording with silence or noise. The language is detected automatically from the audio, and a long quiet or noisy intro can throw it off. Trim it before uploading.
Recording both sides through a phone call on speaker. Phone audio recorded through a laptop or phone speaker introduces significant audio degradation, the signal has already been compressed once by the phone network and then picked up again by a microphone in the room. Whenever possible, record phone interviews using a call recording app rather than holding your phone on speaker.
The transcript itself is only as good as the recording. Get the recording right, and the rest of the workflow takes under ten minutes from upload to clean text file.
Get the text of your own interviews with TranscribTxt for researchers.
Try TranscribTxt free
Free for files up to 30 min: your first one needs no account, a free account gets 3 a week. 99 languages.
First 30 min free · the whole file (up to 3 h) US$2
Try it freeFrequently Asked Questions
How long does it take to transcribe a 60-minute interview?
The AI step is the quick part, so the review pass is what takes time. Budget 15–20 minutes to fix proper nouns and technical terms in a 60-minute interview, compared to 4–6 hours of manual typing. In October 2026, half of the uploaded files were transcribed in 14 seconds or less, and a 145-minute recording took 2 minutes 45 seconds. Times don't include uploading, which depends on your connection.
Is TranscribTxt free for interview transcription?
Yes, for interviews up to 30 minutes. The free plan covers 3 files a week with no credit card and includes TXT and SRT export. For a longer interview, the whole file (up to 3 hours) is $2 once. For several interviews, a one-time $5 Week Pass gives you 7 days of Pro, and Pro ($10/month) removes the per-file limit.
Can I transcribe an interview recorded on my phone?
Yes. Record using your phone's built-in Voice Memos app (iPhone) or Voice Recorder (Android), then upload the M4A or MP3 file to TranscribTxt, from your computer or straight from your phone's browser. A quiet room and a phone placed close to the speaker make the biggest difference to accuracy.
Does TranscribTxt support speaker labels (who said what)?
No. TranscribTxt produces a continuous transcript without speaker attribution on every plan. For a two-person interview, adding names during your review pass is quick because questions and answers alternate; use the SRT timestamps to check any line against the recording.
What file formats can I upload for interview transcription?
TranscribTxt accepts MP3, M4A, WAV, FLAC, OGG, OPUS, AAC, MP4, MOV, WebM, AVI and MKV. Both audio-only and video files work, up to 2 GB per file on every plan, including the free one. If a video is too large, upload just its audio track.