Daytapes

Guides

Glossary

Plain-words definitions for recording, transcripts and the AI around them.

Updated November 14, 2026

Short definitions for the words you'll meet when you record, transcribe and share conversations. They're written for ordinary readers. Nothing here is legal advice. For consent rules, read the sources linked in each entry and the rules of your own state.

All-party consent

A rule that says everyone in a conversation must agree before it's recorded. It's also called two-party consent. States differ on which conversations it covers, so check the rule where you and the other person are. Not legal advice.

Audio file

A file that stores sound. Common types are m4a, mp3, wav, aac, ogg, opus and flac. Daytapes accepts all of these, plus the audio in mp4 and mov files, up to four hours.

Beamforming microphone

A microphone setup that uses several tiny microphones and software to listen in one direction and ignore others. Many speakerphones and some recorders use it to pick out the person speaking.

Bitrate

How much data one second of audio uses, usually in kilobits per second (kbps). A higher bitrate usually means better quality and a bigger file. Speech sounds fine at lower bitrates than music does.

Cardioid

A microphone pickup pattern shaped like a heart. It hears best from the front, less from the sides, and least from behind. It's a good choice when you want one speaker and less room noise.

Chapter

A named section of a long recording, such as "Budget review" at 12:40. Chapters help you jump to the part you need without scrubbing through the whole file.

Clipping

Distortion that happens when a sound is louder than the recorder can handle. Clipped audio sounds harsh and can't be fixed afterward. Lower the input level, or move the microphone farther from loud sources.

Codec

The method used to compress and decompress audio. AAC, MP3 and Opus are codecs. The file type, such as m4a or ogg, is the container that holds the compressed audio.

Data retention

How long a service keeps your data, including after you delete it. Retention rules differ by company, plan and setting. For example, OpenAI's API documentation says abuse monitoring logs are kept for up to 30 days by default (read November 14, 2026).

Diarization

Working out who spoke when in a recording. The result is a transcript where each stretch of speech belongs to a speaker, such as Speaker 1 and Speaker 2. Some tools then let you replace those labels with names.

Dynamic range

The gap between the quietest and loudest sounds in a recording. A recording with too little room for loud moments can clip. A recording that's too quiet can bury speech in noise.

Email-in

A Daytapes feature: each owner gets a private address, <key>@in.daytapes.com, and can send or forward audio there. The key in the address makes it yours, so keep it private.

FLAC

A lossless audio format. It shrinks a file without discarding any sound information, so it sounds the same as the original but takes more space than an MP3.

Hallucination

When an AI states something that isn't in the source. In transcription, it can show up as words that weren't said. In summaries, it can show up as made-up details. Check against the audio.

Hidden parts

A Daytapes sharing feature. When you share a link to a recording, you can hide sections so the people with the link can't see or hear them.

Lavalier microphone

A tiny clip-on microphone, also called a lav or lapel mic. It sits close to the speaker's mouth, so it picks up a clear voice even in a noisy room. Many are omnidirectional.

Lossless and lossy

Lossless formats, such as WAV and FLAC, keep all the audio information. Lossy formats, such as MP3, AAC and Opus, throw away detail the ear is less likely to notice so files get smaller.

M4A

An audio file type used by Apple's Voice Memos and many other apps. It usually holds AAC audio and is a good balance of size and quality for speech.

MCP connector

A way for an AI assistant to reach another service through a standard, approved link. Plaud, for example, describes its MCP connector as a way for AI tools such as Claude and ChatGPT to access Plaud recordings. You grant access once, and you can usually revoke it.

Metadata

Information about a file rather than its content, such as the date, length, device name or location. Metadata can reveal more than you expect, so check it before you share a file.

Mono and stereo

Mono uses one audio channel. Stereo uses two, left and right. Speech recordings are often mono, which halves the file size without hurting how a voice sounds.

Noise reduction

Software or hardware that lowers steady background sounds like hum, fans or hiss. Too much can make voices sound thin or watery. Use the lightest setting that helps.

OAuth

A sign-in method that lets one app access your account on another service without you giving it your password. You approve the access on the service's own screen. Daytapes uses OAuth to connect your Plaud account.

Omnidirectional

A microphone pickup pattern that hears sound equally from all directions. It's handy for a table of people, and it also picks up room noise. Many lavalier mics are omnidirectional.

On-device processing

Work done on your own phone or computer rather than on a company's servers. Local transcription with Whisper is an example. It can keep audio from leaving your device, but it uses your device's power and time.

One-party consent

A rule that says a recording is allowed when at least one person in the conversation agrees, often the person recording. The Reporters Committee says federal law follows this model. State rules can be stricter. Not legal advice.

OPUS

A modern audio codec designed for speech and music at low bitrates. Files often use the .opus or .ogg extension. It's a good choice when you need small files that still sound clear.

Passcode

A short code, usually numbers, that locks something. Daytapes uses a simple passcode for each person's vault. A short passcode is convenient, and it's weaker than a long random password.

Passkey

A way to sign in without a password. Your device holds a secret key, and you approve it with a fingerprint, face scan or device PIN. Passkeys resist phishing because there's no password to type into a fake site.

Real-time transcription

Turning speech into text while the person is still talking. Daytapes shows a live transcript while you record in the app. Live text can be less accurate than a transcript made afterward.

Sample rate

How many times per second a recorder measures sound, in hertz. CD audio uses 44,100 Hz. Speech is often recorded at 16,000 Hz or 48,000 Hz. A higher rate captures higher pitches, but speech doesn't need much.

Signal-to-noise ratio

How loud the voice is compared with the background noise. A higher ratio is better. Moving the microphone closer to the speaker helps more than any software fix.

Speaker label

A tag on a stretch of transcript that says who is talking, either a generic label such as Speaker 1 or a real name. Daytapes shows who's talking by name.

SRT

A common subtitle file type. Each entry has a number, a start and end time, and a line of text. It's one of the transcript files Daytapes accepts, along with .txt, .vtt and .json.

Summary

A short version of a recording that captures the main points. Summaries can miss detail or get it wrong, so use them to find the right moment and then listen.

Talking points

The main topics or statements pulled from a recording, written as a short list. Daytapes creates talking points along with a summary for each recording.

Timestamp

A marker showing when something happened in a recording, such as 14:32. Timestamps let you jump straight to a quote and check it.

Transcript

The written text of a recording. A transcript can be plain text, or timed so each line matches a moment in the audio.

Transcription

The process of turning speech into written text. It can be done by a person or by software. Accuracy depends on audio quality, accents, overlapping speech and jargon.

Vault

A private space for a person's recordings. In Daytapes, each person has a vault protected by a simple passcode.

VTT

WebVTT, a text format for captions and subtitles. A W3C specification defines it, and files begin with the word WEBVTT. Daytapes accepts .vtt transcript files with your audio.

WAV

An uncompressed audio file type. It keeps everything the recorder captured, so files are large. An hour of CD-quality stereo takes roughly 600 MB. Compress it to m4a or opus if you don't need the original.

Watch folder

A folder that a program checks for new files. Daytapes uses a "Daytapes Inbox" folder in a connected Google Drive, Dropbox or OneDrive account. Put audio in it and it becomes a recording page.

Whisper

An open-source speech-to-text model family from OpenAI. Its GitHub page lists the MIT license and model sizes from tiny to large. Many local apps use Whisper or its C++ port, whisper.cpp.

Word error rate

A score for how accurate a transcript is. It counts the words that were wrong, missing or added, and divides by the number of words in the correct text. Lower is better, and results depend heavily on the audio.

Zero data retention

An arrangement where a service doesn't keep your content after processing it. OpenAI's API documentation says it requires prior approval and removes customer content from abuse monitoring logs (read November 14, 2026).