What Is Transcription? Audio & Video Explained

Transcription converts spoken words into written text in the same language — a spoken English interview becomes a written English transcript. Translation converts content from one language to another — a written English document becomes written Spanish. Some services combine both: transcription + translation produces a written transcript in a different language than the original speech.

Verbatim transcription captures every spoken word exactly as said, including filler words (um, uh, like), false starts, repetitions, laughter, pauses, and non-verbal sounds. It’s used in legal proceedings, psychological research, and linguistic studies where the exact manner of speech is important. Most business transcription uses ‘clean verbatim’ instead, which removes fillers while keeping all meaningful content.

Transcription is a text document without timing data — for reading and searching. Captions are timed text synchronized with audio/video, shown in the same language as the spoken content — designed for accessibility (deaf and hard-of-hearing audiences). Subtitles are also timed and synchronized, but typically in a different language than the audio — used for translation. NovaScribe exports all three formats: TXT/DOCX for transcripts, SRT/VTT for captions and subtitles.

AI transcription processes 1 hour of audio in 2–5 minutes. Human transcription typically takes 4–6 hours per hour of audio for a standard turnaround, or 12–24 hours for rush service. The actual time depends on audio quality, number of speakers, and subject complexity. AI services like NovaScribe are near-instant regardless of file length.

Common transcript formats include: TXT (plain text, universally compatible), DOCX (Microsoft Word, most common for editing), PDF (read-only sharing), SRT (SubRip — timed subtitles for video platforms), VTT (WebVTT — web captions for HTML5 video), and JSON (structured data for developers). NovaScribe exports TXT, DOCX, PDF, SRT, and VTT.

Yes — transcription is a key accessibility tool. The Americans with Disabilities Act (ADA) and Web Content Accessibility Guidelines (WCAG 2.1 AA) require that audio and video content be accessible to deaf and hard-of-hearing users. Transcripts and captions fulfill this requirement. Universities, government agencies, and companies subject to Section 508 compliance must provide transcripts or captions for all recorded audio/video content.

AI transcription reaches 95–98% accuracy in ideal conditions — clear audio, single speaker, standard accent, general vocabulary. In challenging conditions (multiple speakers, background noise, heavy accents, technical jargon), accuracy typically falls to 70–90%. For most business use cases like meeting notes, podcast show notes, and YouTube captions, AI accuracy is more than sufficient.

Dictation is the real-time process of speaking for immediate capture — like speaking to a voice assistant or dictating a letter. Transcription is the conversion of pre-recorded audio into text after the fact. The key difference is timing: dictation happens live, transcription happens later. Many AI transcription tools can also handle dictation (real-time speech-to-text), but the primary use case is post-recording conversion.