How to Transcribe a Phone Interview: A Step-by-Step Guide

Record the call with a phone call-recording app, then upload the audio file to an AI transcription tool to get text. Expect lower accuracy than studio audio because phone lines compress sound: budget for roughly 88 to 94 percent on a typical call, then proofread names and key quotes before you publish.

Phone interviews are everywhere in journalism, research, and recruiting, but the audio is harder to transcribe than a face-to-face conversation. The phone network strips out part of the frequency range your voice normally occupies, and that missing detail is exactly what a transcription engine relies on. The good news: with a clean recording and the right workflow, you can still get a usable, accurate transcript in minutes. Here is how to do it end to end.

Step 1: Record the call cleanly

Everything downstream depends on the recording. A muddy file is the single biggest cause of bad transcripts, so spend a minute getting this right.

1. Pick a recording method. On iPhone, recent iOS versions can record calls natively, or you can use a dedicated call-recording app. On Android, several call-recording apps capture both sides of the line. If app-based recording is unreliable on your device, put the call on speakerphone and record with a second phone or a voice recorder placed close to the speaker.
2. Prefer a direct line over speakerphone. A direct in-app recording captures each side cleanly. Speakerphone adds room echo, picks up background noise, and can blur the two voices together, which makes speaker separation harder later.
3. Find a quiet room. Close windows, turn off fans and notifications, and avoid coffee shops. Background noise competes with speech and drags accuracy down.
4. Test for 10 seconds first. Record a short clip, play it back, and confirm both voices are audible before the real interview starts.
5. Start recording early. Hit record before the conversation begins so you never lose the opening question or an unexpected early answer.

Save the file in a common format. Most call-recording apps export MP3 or M4A, both of which are ready to upload.

Step 2: Get consent before you record

Recording laws vary. Some places allow recording when just one party consents; others require every party on the call to agree. Rules differ by state and country, and the location of each person on the line can matter.

The safe habit: ask for permission at the start of the recording and capture the "yes" on tape. That protects you and creates a clear record. If you are unsure what applies to your situation, check the rules for your jurisdiction before you record. This is general guidance, not legal advice.

Step 3: Upload and transcribe

Once you have a clean file, the transcription itself is fast.

1. Upload your audio. On TranscribTxt, drag in the recording. Supported inputs include MP3, M4A, WAV, MP4, MOV, and WebM, so most call-recording exports work directly.
2. Let the model detect the language. TranscribTxt runs on OpenAI's Whisper large-v3 turbo and auto-detects across 99 languages, so there is nothing to set.
3. Mark the speakers. TranscribTxt doesn't label speakers, so the transcript comes back as continuous text. In a two-person interview, questions and answers alternate, so marking interviewer and source during review is quick.
4. Review the draft. The transcript appears as punctuated text. Skim it against the audio, focusing on the moments that matter most; the SRT download gives you a timestamp on every line.
5. Export in the format you need. Download as TXT for a clean reading copy, SRT if you need timed captions, or JSON with word-level timestamps (Pro) if you are feeding the data into another tool.

Your uploaded audio is deleted after processing, and the transcript isn't stored on our servers, which matters when the recording contains a private conversation.

If you want the full general workflow beyond phone calls, see our guide to transcribing any interview.

Step 4: Understand realistic accuracy on phone audio

Set expectations before you start editing. Clean, close-mic studio audio commonly transcribes at around 95 to 98 percent accuracy. Phone and other compressed audio typically lands closer to 88 to 94 percent. The drop is not the tool failing; it is physics.

Three things work against a phone recording:

• Narrow frequency range. Phone lines transmit a limited slice of the audio spectrum, so consonants and subtle sounds the engine uses to distinguish words are simply missing.
• Compression artifacts. Cellular and VoIP codecs squeeze the signal, introducing distortion that smears word boundaries.
• Dropped packets and lag. A spotty connection produces gaps and glitches that read as garbled words in the transcript.

You can push toward the top of that range by recording over a direct line, using a quiet room, and asking your source to avoid talking over you. For a deeper breakdown of what drives the numbers, read our AI transcription accuracy guide.

Step 5: Fix names, quotes, and terms

The proofreading pass is where a phone transcript goes from "draft" to "publishable." Focus your time where errors are most likely and most costly.

1. Verify proper nouns first. Names, companies, places, and product names are the most common errors because the model cannot guess spelling from compressed audio. Check each against your notes or a quick search.
2. Re-listen to key quotes. Before you put anything in quotation marks, play that segment back and confirm the wording word for word. A single transcription slip can change the meaning of a quote.
3. Use Find and Replace for repeat terms. If your source uses a recurring acronym or technical term the engine misheard, fix every instance in one pass.
4. Mark uncertain spots. Where the audio was unintelligible, leave a clear marker like [inaudible] rather than guessing, so you can follow up if needed.
5. Keep the timestamps. They let you jump straight back to any moment in the recording during fact-checking.

For reporters working under deadline, our notes on transcription for journalism cover quote accuracy and sourcing in more depth, and our roundup of the best interview transcription software for journalists compares options for newsroom workflows.

Quick checklist

• Record over a direct line, in a quiet room, starting before the call begins.
• Confirm consent on the recording and check your local rules.
• Upload the MP3 or M4A, mark the speakers, and review against the SRT timestamps.
• Expect roughly 88 to 94 percent accuracy and proofread accordingly.
• Verify every name and re-listen to every quote before publishing.

Try it on your next call

You can transcribe a phone interview for free on TranscribTxt: your first file needs no account, and the Free plan covers 3 files a day, up to 30 minutes each, with no card required. For longer calls or regular interviews, Pro is $10 per month with no per-file length cap, or a one-time $5 Week Pass gives you 7 days of Pro. Record cleanly, upload, and let the AI handle the first draft so you can spend your time on the words that matter.