How to Get Audio Transcripts

An audio transcript is a written version of what was said in a recording, meeting or video call. You can get one from a transcription feature in an app you already use, from a dedicated speech-to-text tool, or by typing it yourself. This guide covers preparing the audio, generating the text and checking it before you rely on it.

Before you start

First, decide whether you need to transcribe an existing file or a live conversation. Then find out where the speech is processed. Some tools convert speech to text on your own device. Others upload the audio to an online service. The tool's documentation or privacy settings should say which one applies, and whether recordings or transcripts are kept and for how long. If this is not stated clearly, assume the audio leaves your device and handle sensitive material accordingly.

• Consent from everyone being recorded, where the law or your organization's rules require it.
• A recording in a file type the tool accepts. Check its supported formats and any limits on file size or duration.
• For live transcription, the correct input device selected in the app and microphone access allowed in your operating system.
• The spoken language of the recording, plus the number of speakers if the tool asks for it.
• Enough remaining allowance on your account if the service limits minutes or uploads.

Step-by-step instructions

Menu names and option locations differ between apps and versions. Use the steps below as the order of work, and follow the labels your tool actually shows.

• Record clean audio. Use a quiet room, keep the microphone at a steady distance from the speaker, and switch off fans, alerts and other sources of background noise.
• Set the input level. Speech should register clearly on the level meter without hitting the maximum or triggering a clipping warning. Distorted peaks and very quiet speech both reduce accuracy.
• Keep the original. Make a copy before you convert, trim or clean up the file, so you can always return to the untouched recording.
• Convert only if necessary. If the tool rejects your file, export a copy in a format from its supported list. Use the sample rate the tool recommends. Lowering the rate will not improve recognition and can remove detail.
• Choose a transcription method: the built-in feature of your recorder or call app, a standalone speech-to-text app or service, or manual typing for short or sensitive clips.
• Select the spoken language, and turn on speaker labels or timestamps if the tool offers them and you need them.
• Start the transcription and let it finish. Long files can take a while. If processing runs on your device, keep the app open until it completes.
• Export the transcript in the format you need. Common choices are plain text, a document, or a subtitle file with timestamps, if the tool supports them.

In video calls, transcription may be limited to the meeting host or controlled by an administrator, and some plans or versions do not include it at all. If you cannot find the option, ask the host or check the platform's help pages for your account type. Other participants may be notified when transcription starts, so tell them in advance.

Check the output

Treat an automatic transcript as a draft. Play the recording while you read along. Pay closest attention to names, numbers, dates and technical terms, because speech recognition most often gets these wrong.

• Timestamps line up with the audio when you jump to several points in the file.
• Speaker labels match the people talking. Overlapping speech is often merged or assigned to the wrong person.
• Quiet passages and sections with music or noise are not missing, and they have not been filled with words nobody said.
• The exported file opens correctly in the program you will use, and characters for your language display properly.
• You know where the transcript is stored and who can access it, especially if an online service kept its own copy.

Accuracy depends on recording quality, accents, crosstalk and the tool itself. If the text will be quoted, published or kept as an official record, have a person review it against the audio.

Common problems and next steps

Empty or very short transcript: play the source file to confirm it actually contains speech. A recording made with the wrong microphone selected, or with the input muted, can be silent. For live sessions, confirm the right input device is active and the app has microphone permission.

Only one side of a call appears: check which sources the recording app was set to capture. Your voice comes from the input device, and remote participants come through the output device. Some apps record these separately or capture only one of them.

Many recognition errors: the usual causes are background noise, a low input level or a microphone placed too far away. Re-recording is the most reliable fix. If that is not possible, apply noise reduction to a copy, transcribe both versions and compare them, since heavy processing can add artifacts that make recognition worse.

Wrong language or garbled text: confirm the language setting matches the recording. For recordings that mix languages, check whether your tool supports more than one language per file. If it does not, transcribe each part separately.

File rejected or upload fails: compare the file type, size and length with the tool's stated limits. Export a supported format from your copy, or split a long recording into shorter parts.

Once the transcript is corrected, store it together with the original audio so you can check disputed passages later. If you used an online service and no longer need its copies, review its retention settings and delete the files there.