Skip to content
KolWrite Open KolWrite

Guides and research

Speech, language, and the work that starts after the transcript is done.

Articles on speech recognition, multilingual audio and accessibility, and on the choices that turn a recording into text you can rely on.

Diagram of the transcription pipeline: a sound wave is cut into short windows, an encoder turns the windows into a representation of speech, a decoder scores candidate word sequences, and separate stages add speaker labels and timestamps to produce a timed transcript.

Speech recognition

How a recording becomes text you can work with

Transcription systems estimate words rather than hearing them. We'll trace the chain from a sound wave to a timed transcript and look at the places where it tends to break.

6 min read

Research

An example of how transcription errors are counted: a human reference line sits above the machine output, showing one substituted word, one deleted word, and one inserted word, with the total divided by the number of reference words.

Evaluation

Transcription accuracy cannot be boiled down to one number

Word error rate is a real metric, though it is easy to quote without the conditions behind it. Even with a strong score, a transcript can still fail at the job you bought it for. This article covers what that number counts and what we measure instead.

6 min read

A caption timeline showing how long each caption stays on screen compared to the time needed to read it, with one overly long caption flagged next to a shorter version that fits its time slot.

Accessibility

Captions are part of the content

A transcript, a translated subtitle track, and a caption file can contain the exact same words while serving three distinct purposes. We explain what each format does, what the standard requires, and how we check media before it goes live.

6 min read

Practical guides

Recording and audio files

The export dialog with format choices, speaker names and timestamps.

Recording

Which audio file format should you use for transcription?

File format matters less than you might think, and how a file is saved matters more than the extension on it. We compare common formats, look at what compression costs you, and explain which file to keep.

6 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Recording

How to record a phone call, and what to check first

Phones make recording a call harder than it needs to be. We cover your options, how to get clear audio from each one, and the consent questions to settle before you start.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Recording

How to record the sound your Mac plays

A room microphone won't cleanly pick up a call, a webinar, or a lecture playing in your browser on a Mac. You can fix that by recording the computer's own audio directly, as long as you set it up carefully.

6 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Recording

How to transcribe a recording that lasts for hours

Multi-hour recordings need a different approach to uploading, navigating, and reviewing than short clips. A little preparation upfront saves hours of work later.

7 min read

Phones, computers and watches

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Devices

How to transcribe a recording on your iPhone

Your iPhone already records good audio. Depending on where that audio is stored, you have three ways to turn it into a written transcript.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Devices

How to transcribe audio and video on a Mac

Reviewing a transcript is comfortable on a Mac because you have a large screen, a physical keyboard, and enough space to keep your audio, text, and exports open side by side. Here is a simple way to go from an audio file to a finished document.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Devices

How to turn voice memos into text you can use

Voice memos are quick to record when you have a thought, and a pain to dig through later. Transcribing those recordings gives you text you can read, search, and share.

7 min read

Hebrew, Arabic and other languages

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Languages

Transcribing speech that moves between Hebrew and English

In Israeli meetings, lectures, and interviews, people often switch languages mid-sentence. That trips up transcription software. Here is why it happens, and how to get a transcript you can rely on.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Languages

Transcribing Russian and other languages used in Israeli workplaces

Israeli teams, clinics, classrooms, and call centers work in several languages besides Hebrew. Here is how to get a good transcript from Russian, French, English, or mixed-language audio, and what to check before you share it.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Translation

How to translate an Arabic recording into Hebrew

Because Arabic and Hebrew are related and read in the same direction, translating between them looks easier than it is. The hard part is sorting out dialect, names, numbers, and the months of the year.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Translation

How to translate a Hebrew recording into English

A reliable approach has two steps. First, capture the Hebrew accurately as text, and then translate it. Each stage needs its own set of checks. Skipping that first step is the main reason an English translation goes wrong.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Languages

Why spoken Hebrew is hard to turn into text

Spoken Hebrew looks simple enough on the page, even though the script drops vowel sounds, attaches prefixes directly to words, and has no fixed spelling for everyday vocabulary. All three shape where transcripts make mistakes.

7 min read

Meetings

The speaker panel next to a transcript, with names and segment counts.

Meetings

Turning a transcript into action items with owners, dates, and follow-up

Good intentions at the end of a meeting fade within a week unless you have clear action items. Here is how to pull them out of a transcript, write them so someone can actually finish the work, and confirm them before people start relying on them.

8 min read

The speaker panel next to a transcript, with names and segment counts.

Meetings

How to get a meeting summary you can trust

A summary is only worth sending if people who missed the meeting can act on it. We cover what belongs in one, how to ask an assistant for it, and what to double-check before hitting send.

8 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Meetings

How to document meetings so the whole team can catch up

On a remote team, the people who need to know what happened in a meeting often cannot attend it. Writing things down consistently turns those conversations into a searchable record instead of something people have to remember.

7 min read

The speaker panel next to a transcript, with names and segment counts.

Meetings

Speaker identification: how a transcript knows who is talking

Without speaker labels, a transcript is just a block of text. Adding them turns it into a conversation you can actually quote. Here is how software creates those labels, why it sometimes mixes people up, and how to fix the mistakes.

8 min read

The speaker panel next to a transcript, with names and segment counts.

Meetings

How to transcribe a Google Meet recording

Google Meet saves recordings straight to Google Drive. Getting a verified transcript from that file only takes a few steps. The hard part is usually having permission to record and tracking the file down afterward.

7 min read

The speaker panel next to a transcript, with names and segment counts.

Meetings

How to transcribe a Microsoft Teams recording

Teams saves meeting recordings to your organization's Microsoft storage rather than inside the app. Once you know where the file lands, who can open it, and how long it stays there, everything else is straightforward.

7 min read

The speaker panel next to a transcript, with names and segment counts.

Meetings

How to transcribe a Zoom recording

Zoom only records a meeting if someone hits record, and once the call ends you have to know where that file goes. Here is where to find it, which file to choose, and how to turn it into a transcript.

8 min read

By profession

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Healthcare

Dictating clinical notes: protect the patient and check the details

Many clinicians dictate notes because speaking is faster than typing. That only saves time if you protect patient privacy first and verify every critical detail. This article gives general guidance, not medical, legal, or compliance advice.

7 min read

The speaker panel next to a transcript, with names and segment counts.

Research

Transcribing a focus group: how to tell many voices apart

A focus group usually puts six to ten people around one table, and they talk over each other. Your transcript is only as good as its speaker labels. Here is how to set up the room, record the session, and label, check, and analyze the transcript.

8 min read

The speaker panel next to a transcript, with names and segment counts.

Media

Interview transcripts for journalists: accurate quotes and protected sources

Putting someone in quotation marks means promising your readers and your source that the words are exact. A transcript helps you locate a quote quickly, while the recording proves it was actually said. Below is a practical workflow for checking quotes and protecting your sources.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Students

Turning a lecture recording into study notes you can actually use for exams

Recording a lecture only helps if you can find things in it later. Transcribing the audio lets you search it, and a few simple habits will help you turn that text into proper study notes. This guide covers getting permission to record, writing clear notes, and using them to study.

7 min read

The speaker panel next to a transcript, with names and segment counts.

Legal

Transcribing recorded evidence and hearings: a verification checklist

In legal work, the recording itself is the evidence. A transcript is just a way to work with it. Here is a checklist for keeping the original file intact, checking the text, labeling speakers, and handling confidential material. This is general guidance, not legal advice.

8 min read

The speaker panel next to a transcript, with names and segment counts.

Research

Transcribing research interviews: consent, anonymity, and method

In qualitative research, your transcript is your data. Here is what to decide before your first interview, how to transcribe and check the files, how to protect your participants, and how to get the text ready for analysis.

8 min read

Subtitles and exports

The export dialog with format choices, speaker names and timestamps.

Subtitles

Burned-in captions vs closed captions: how to choose

You can render captions straight into your video or supply them as a separate track. That decision affects who sees the text, how you fix typos, and where the file can play.

7 min read

The export dialog with format choices, speaker names and timestamps.

Subtitles

Why right-to-left subtitles break, and how to fix them

You can have all the right words in a Hebrew or Arabic subtitle and still watch it display incorrectly. A period lands on the wrong side, a number flips around, or an English name splits a line apart. All of these problems have the same cause, and you can fix it.

7 min read

The export dialog with format choices, speaker names and timestamps.

Subtitles

SRT vs. VTT: Which subtitle format should you use?

Both formats are plain-text files that match words to timestamps. They look almost the same on the surface, but web players handle them differently, and they differ in how they support styling and how a player reads the file.

7 min read

The export dialog with format choices, speaker names and timestamps.

Subtitles

Subtitle rules: line length, reading speed, and timing

Good subtitles are easy to read because they follow a few measurable habits: keep lines short, leave them on screen long enough, and break lines where the meaning breaks. Here are the standard rules and why they work.

6 min read

The export dialog with format choices, speaker names and timestamps.

Video

How to turn a video file into text and subtitles

Getting a checked transcript and subtitle file from an MP4 or MOV video only takes a few steps. Since the words come from the audio track, a quick sound check at the start prevents most problems.

6 min read

The export dialog with format choices, speaker names and timestamps.

Documents

How to move a transcript into Google Docs

Google Docs is an easy place to review a transcript with other people. Moving the text over only takes a few steps, though right-to-left languages need a quick adjustment.

7 min read

The export dialog with format choices, speaker names and timestamps.

Documents

How to turn a transcript into a clean Word document

A raw transcript is just the starting point. You usually take it into Word to turn it into meeting minutes, an interview write-up, or a report. Here is how to export the file, handle the right-to-left settings for Hebrew and Arabic, and format the document so it is easy to read.

7 min read

Choosing and using transcription

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Choosing a tool

Dictation vs. transcription: two ways to turn speech into text

Both start with speech and end with text, though they are built for different jobs. Dictation lets you write by talking out loud. Transcription turns an existing recording into something you can read.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Developers

A practical guide to using a speech-to-text API

Adding transcription to a product comes down to integration and testing. Here is what a speech-to-text API takes in and sends back, how to work with word timings and speaker labels, and how to test and secure your setup.

8 min read

The speaker panel next to a transcript, with names and segment counts.

Choosing

How to test a transcription tool with your own worst audio

A vendor's demo shows how a tool works when everything goes right. Your own recordings show what it actually does. Here is how to run a short, fair test, pick the right things to measure, and read the results without getting fooled by a single accuracy score.

7 min read

The KolWrite transcript view: speaker names, timestamps and the line currently playing.

Privacy

A privacy checklist for transcribing recordings

Recordings are some of the most sensitive files you own. Before uploading one to any service, check where it goes, who can access it, whether it is used to train a model, and how you remove it.

6 min read