How to Transcribe Cassette Tapes and Old Recordings

Identify what you actually have

Old speech recordings arrive in a handful of formats, and each needs a different machine and setting. Getting this wrong is the most common way to produce a poor digital copy from a perfectly good tape.

Compact cassette: Plays at 4.76 cm/s (1 7/8 inches per second). Common lengths are C60, C90 and C120, holding about 30, 45 and 60 minutes per side.
Microcassette: Used in dictation machines and answering machines, usually at 2.4 cm/s, with many machines offering a half-speed 1.2 cm/s setting. Needs a microcassette player set to the same speed used for recording.
Mini-cassette: A different, smaller dictation format that won't play in a microcassette machine, despite the similar name.
Reel-to-reel: Common speeds are 3¾ and 7½ inches per second, with 1 7/8 on some home machines and 15 for studio work. Tapes can be full-track, half-track or quarter-track, mono or stereo, and some were recorded in both directions.
VHS: Audio can be on the linear track along the tape edge or on the Hi-Fi stereo tracks if the recorder supported them. Tapes recorded at the slower LP or EP speeds have noticeably poorer linear audio.

Write down what is on the label and the box insert, and photograph them before you start. Labels fall off, and the handwritten note "Grandpa, Easter, side B continues" is part of the record.

Playback equipment and condition

The deck matters more than the computer. A well-maintained used cassette deck from a hi-fi system usually outperforms a cheap new USB cassette player, whose transport and converters are often basic. A cheap USB player is acceptable for a quick listening copy; for anything you want to keep, use a decent deck and a separate audio interface.

Before playing anything you care about:

Clean the heads, capstan and pinch roller with isopropyl alcohol and a cotton swab. Oxide buildup dulls the sound and can damage tape.
Play a tape you don't care about first, to confirm the deck runs at a steady speed and doesn't chew tape.
Check the azimuth, the angle of the playback head. A misaligned head makes a tape sound dull and muffled. Many decks have a small adjustment screw; adjusting while listening for the brightest, clearest sound is a common technique, but if you aren't comfortable doing it, ask a repair shop.
Match the noise reduction. Many cassettes were recorded with Dolby B. If a tape sounds unnaturally bright and hissy with noise reduction off, try it on; if speech sounds dull and lifeless with it on, try it off. Keep whichever sounds natural, and note the setting.
Inspect old or fragile tapes before playing them. Some reel tapes from the 1970s and 1980s suffer from binder breakdown, often called sticky-shed syndrome, and squeal or shed oxide when played. Moldy, brittle or sticky tapes can be destroyed by playback; send irreplaceable ones to a professional digitization service.

Capture settings that keep options open

Connect the deck's line output to the line input of an audio interface or a recorder, not a microphone input and preferably not a headphone jack, which adds the deck's headphone amplifier to the signal. For VHS, use the VCR's audio outputs, or capture the video and audio together with a capture device if you want the pictures too.

Record at 24-bit, at 48 kHz or higher. Archival programs often specify higher rates for preservation masters, a topic covered in oral history transcription, but for transcription the rate is not critical, because recognition runs at 16 kHz. Set levels so the loudest passage peaks around -6 dBFS; tape levels vary within a recording, and you need headroom for the loud moments.

Capture each side as one continuous file, without processing. Leave noise reduction, EQ and normalizing for later, on a copy. If a mono cassette plays as two identical channels, capture in stereo anyway; it costs nothing, and you can fold it to mono afterward. For reel tapes recorded in both directions, the second direction needs either a turned-over tape or a deck with auto-reverse heads, and it is easy to miss half the content if you don't check.

Hiss, wow and flutter: what to clean and what to leave

Tape hiss is a steady high-frequency noise. Light noise reduction on the transcription copy can make listening less tiring, but heavy settings remove consonants along with the hiss and can produce watery artifacts that confuse recognition more than the hiss did. The trade-offs are covered in speech recognition and background noise; mains hum from old equipment is covered in audio hum and buzz.

Wow and flutter are speed variations: wow is a slow wavering of pitch, flutter a faster trembling. They usually come from worn belts, a dirty capstan or stretched tape. Mild wow doesn't trouble a recognizer much, but severe wavering makes voices sound unstable and lowers accuracy. Replacing the deck's belt often fixes wow at the source. Specialist restoration software can reduce wow and flutter after capture, but it is a job for patience and good ears.

Speed errors are easier. If voices sound high and fast or low and slow throughout, the tape was probably recorded at a different speed from the one used for playback, which is common with microcassettes and home reel machines. Re-capture at the correct speed if the deck allows it; otherwise correct it in an audio editor by changing speed with pitch following, until a known voice sounds right.

A worked example: a box of family cassettes

Example

A hypothetical family has fourteen cassettes of a grandmother's stories and a microcassette from her shop's answering machine. They clean and test a borrowed hi-fi deck, photograph every label, and capture each side as a 24-bit, 48 kHz WAV through a small audio interface. One tape sounds dull until the azimuth is adjusted; another plays hissy and bright until Dolby B is switched on. The microcassette plays too fast until the player is set to its half-speed setting. A full C90 tape, two 45-minute sides, costs 90 credits (9¢) to transcribe, and the family translates the transcripts for relatives abroad.

The most useful habit in that workflow is boring: one file per tape side, named consistently, with a log entry written at the time of capture.

Digitizing and transcribing, step by step

Sort the tapes by format, photograph labels and inserts, and give each tape an identifier such as FAM-07.
Clean the deck, test it on a disposable tape, and check speed, azimuth and noise reduction on each new tape.
Connect line out to line in and set peaks around -6 dBFS on the loudest passage you can find.
Capture each side as one uncompressed WAV or FLAC file, named with the identifier and side, for example FAM-07_sideA.wav.
Log the tape label text, the deck settings and anything odd you heard during capture.
Keep the raw capture untouched and make a copy for transcription, applying only gentle, tested cleanup.
Transcribe the copy, one side per file, keeping each file under 2 hours.
Review against the audio, especially names, places and dates, and note tape faults such as dropouts as [tape damage] with a timestamp.

Mistakes that cost you the recording

Playing a fragile or moldy tape on a home deck. A single pass can shed the oxide that holds the sound; professionals have tools and treatments for damaged tape.
Fast-forwarding and rewinding old tapes repeatedly, which stresses splices and leaders. Play straight through when you can.
Recording through a microphone placed in front of the deck's speaker. It adds room noise and loses the clarity of a direct cable connection.
Saving only an MP3 made with heavy noise reduction. Once the raw capture is gone, you can't undo the processing.
Assuming the end of side A is the end of the recording. Speech often continues on side B, or after a long gap on the same side.

Where mydubly fits for digitized tapes

mydubly starts after digitizing: it doesn't play tapes or capture audio. Upload the digitized side as WAV, FLAC, MP3, M4A, AAC or OGG to audio to text, or a captured VHS file as MP4, MOV, MKV, WebM or M4V to video to text. Most capture setups produce WAV, and WAV to text covers that format specifically. Each file can be up to 2 hours, so a long VHS recording may need to be split first; transcribing long audio files covers splitting at natural pauses.

The browser converts the audio to 16 kHz mono and sends compressed audio chunks over HTTPS for transcription with Whisper; a captured video file stays on your device. You get a plain transcript, a timestamped transcript and SRT and VTT subtitles, plus a translated transcript in one of the 21 supported languages if you choose one. There are no speaker labels and no tape restoration, so speaker names and any cleanup are your work. Transcription costs 1 credit per minute, so a 30-minute cassette side costs 30 credits (3¢).

Next step: digitize one side well

Start with a single tape that matters but isn't fragile. Clean the deck, check the settings, capture one side, and transcribe it. Review the result against the audio; if the errors are mostly names and places, your capture is good and the rest of the collection can follow the same routine. If you are building a lasting archive of interviews, read oral history transcription for the consent, metadata and preservation side.

Frequently asked questions

Should I digitize tapes myself or use a service?

Do it yourself if the tapes are in good condition, you have or can borrow a decent deck, and you have the time; a 90-minute cassette takes 90 minutes to capture. Use a professional service for fragile, moldy, sticky or broken tapes, unusual formats you can't play, or recordings that are irreplaceable. Ask any service what file format and resolution they deliver, and request uncompressed WAV.

What if the tape snaps or tangles?

Stop the machine immediately and don't pull the tape out by force. Cassette tape can often be spliced with proper splicing tape, and a cassette shell can be replaced by transferring the reels to a new one. For valuable recordings, take the damaged tape to a professional rather than attempting a repair, because careless handling can crease or stretch the tape at the damaged point.

Can I transcribe a tape without digitizing it first?

Not with a transcription tool, which needs a file. You could play the tape near a phone and record it, but that adds room noise and loses clarity, so the transcript will be worse. A direct cable connection from the deck to a computer or recorder gives a much cleaner file for little extra effort.

Is a high sample rate worth it if I only want a transcript?

Not for the transcript itself, because recognition runs on 16 kHz audio. Higher rates matter for preservation, so if the tape might be the last copy of a recording, capture at archival settings once and make smaller copies from that file. Re-playing old tapes to capture them again adds wear, which is why it pays to capture well the first time.

How do I deal with a tape that has two languages on it?

Transcribe it once and check the result carefully. mydubly detects the spoken language automatically, but recordings that switch languages, common in family and community recordings, can produce sections in the wrong language or partly translated text. Translating mixed-language videos explains the behavior and practical workarounds such as splitting the file at language changes.