Interview Transcription Online

From recording to quotable text

Whether it is a research interview for a thesis, a source for an article or a user interview for a product, the recording is only useful once it is text: coded, quoted, searched. Transcribing by hand takes four to six hours per hour of audio. This page shows what the automatic version looks like on your own recording, using the first 5 minutes, before you commit the rest to the app.

Who said what

Interviews are two-voice recordings, and the transcript reflects that: each turn is labelled Speaker A or Speaker B with a timestamp, so the interviewer's questions and the participant's answers stay apart. Three-person panels work the same way. The labels are consistent across the transcript — Speaker B at minute one is Speaker B at minute four — which is what makes coding by respondent possible.

What the preview does

Your browser reads the file and keeps the first 5 minutes, cutting longer recordings locally and re-encoding that part as 16 kHz mono WAV so the rest is never uploaded. The audio goes over an encrypted connection to the transcription engine behind the SozAI app (AssemblyAI). The language is detected automatically across 99 languages, punctuation is added, and filler words are lightly cleaned — a readable transcript rather than a strict verbatim one. When the text is delivered, the audio and transcript are deleted at the provider.

Formats and limits

• M4A, MP3, WAV, AAC, FLAC, OGG/OPUS, WEBM, up to 25 MB as-is; longer recordings are trimmed to 5 minutes in the browser first (up to 30 MB on a computer, 12 MB on a phone).
• Three previews per device per day, no account.
• Phone calls recorded through a speaker and interviews over a café table both transcribe, with more errors on the quieter voice.

Consent and confidentiality

Recording a person requires their consent almost everywhere, and research ethics boards usually require it in writing. Transcribing does not change that. If your protocol restricts where audio may be processed, note that the preview is processed by a US-based provider unless the site has been set to its EU endpoint, and use the app's project settings for anything covered by a data agreement.

The whole interview

An hour-long interview is twelve previews' worth of audio. The SozAI app transcribes it in one go, keeps the speaker labels, produces a summary of themes, and stores every interview in a library you can search across. Import from Voice Memos, Files or a shared recording. The first 30 minutes are free.