How to Transcribe an Audio File (Speech to Text)

Transcription used to be tedious manual work or expensive professional services. AI has made it nearly free and very accurate. Whether you’re transcribing interviews, lectures, podcasts, or voice notes, here are the practical options.

The fastest path: Otter.ai

Otter.ai (free tier, paid for heavy use):

Sign up for free account
Upload audio (or record directly)
Otter generates a transcript with timestamps and speaker labels
Review and export as TXT, DOCX, PDF, SRT

Free tier: 300 minutes/month, decent quality.

For most casual transcription needs, Otter is the right answer.

OpenAI Whisper (free, open source)

Whisper is OpenAI’s open-source transcription model. It runs locally on your computer:

Install on Mac/Linux:

pip install -U openai-whisper
brew install ffmpeg  # if not already

Run:

whisper audio.mp3 --language en --output_format srt

Generates a .srt subtitle file (also .txt and other formats).

Quality: comparable to commercial services. With the “large” model, often better than humans for clean audio.

Speed: depends on your computer; 1-5× real-time on consumer hardware.

Cost: free; runs entirely locally.

For sensitive content (legal, medical, business meetings), Whisper running locally is the privacy-friendly choice.

Apple Dictation

Built into Mac and iOS:

Mac: System Settings → Keyboard → Dictation. Press the dictation key (or Fn key twice) → speak into the mic → text appears in any text field.

iOS: tap the mic icon on the keyboard → speak → text appears.

Limitations:

Designed for live dictation, not recorded audio transcription
For recorded audio: play through speakers while dictating (works but imperfect)
Accuracy is decent for clear speech

Google Recorder (Pixel phones)

If you have a Pixel phone:

Open Recorder app
Record (or import audio)
Auto-transcribed with speaker labels and search

The transcription is on-device, doesn’t require internet, and is very fast. One of the best built-in transcription experiences.

Trint, Rev, Descript (paid)

For professional transcription:

Trint: ~$60/month for many features.

Rev: $1.50/minute for human transcription, $0.25/minute for AI.

Descript: $12-30/month for editing-integrated transcription. Excellent for podcast editors.

For one-off projects: pay-as-you-go services. For ongoing work: subscription.

Free vs paid comparison

Service	Cost	Quality	Speed	Best for
Otter.ai free	Free (300 min/mo)	Good	Fast	Casual use
Whisper (local)	Free	Excellent	Medium	Privacy-sensitive
Apple Dictation	Free	Decent	Real-time	Quick notes
Google Recorder	Free	Excellent	Real-time	Pixel users
Rev (AI)	$0.25/min	Very good	Hours	Pay-per-use
Rev (human)	$1.50/min	Excellent	Days	Professional
Descript	$12-30/mo	Excellent	Fast	Podcast workflow

Preparing audio for transcription

Better audio = better transcription:

Clear audio:

Single speaker at a time
Close to microphone
Minimal background noise
Speakers don’t speak over each other

Common issues that hurt quality:

Multiple speakers in one room
Background music or noise
Poor mic quality
Strong accents (less of an issue with modern models)
Heavy technical jargon
Echo / reverberant rooms

For best results: record with attention to audio quality, then transcribe.

File format

Most transcription services accept:

MP3 — most common
WAV — high quality
M4A / AAC — Apple/iPhone recordings
MP4 (video with audio) — many services extract audio automatically
FLAC, OGG — supported by most

If your audio is in an unusual format: convert with WAV to MP3, FLAC to MP3, MP4 to MP3, etc.

Post-processing the transcript

Even AI transcripts need review:

Speaker labels: AI guesses; verify and assign actual names
Proper nouns: names, places, technical terms often misspelled
Punctuation: usually decent but check for missing periods/commas
Timestamps: useful for navigating long transcripts
Filler words: “um”, “uh”, “you know” — keep or remove depending on use

For published transcripts (interviews, articles): edit for readability. For internal notes: rough transcript is fine.

Subtitle generation

Many transcription services export SRT (subtitle file) format:

Transcribe audio
Export as SRT
Pair with video → see Add Subtitles to a Video

This is the standard workflow for adding captions to videos.

Privacy considerations

Cloud services (Otter, Rev, Trint, etc.): your audio is uploaded to their servers. For:

Casual content: fine
Personal conversations: consider whether you’d be comfortable with the service having that recording
Business confidential: review their privacy/security claims
Legal/medical: stick to local-only tools

Local tools (Whisper, Google Recorder on-device): nothing uploads. Privacy-preserving.

For sensitive content, Whisper running locally is the right answer.

Use cases

Interview transcription: Otter or Rev (AI) for fast, paid Rev (human) for legal quality.

Podcast episode transcripts: Descript fits the podcast workflow.

Meeting notes: Otter integrates with Zoom/Google Meet for live transcripts.

Lecture notes: Otter, or Google Recorder if on Pixel.

Voice memo to text: Apple Dictation, Otter, or Google Recorder.

Legal/medical: Whisper locally for privacy, or specialized HIPAA-compliant services.

After transcribing

Trim corresponding audio: Audio Trimmer for the audio file
Save as multiple formats: most services export TXT, DOCX, PDF, SRT
Edit and publish: many publishing workflows start with transcripts

TL;DR

Easiest: Otter.ai free tier (300 min/month)
Privacy-friendly: OpenAI Whisper running locally
Pixel phone: Google Recorder is excellent and on-device
For paid quality: Rev (AI or human), Descript (podcast workflow)
Audio quality matters: clear input = clean output
Review the transcript: AI is good but not perfect
For subtitles: export SRT and combine with Add Subtitles to Video