Transcription guides
In-depth, practical guides to getting accurate transcripts from your recordings.
Diarization separates voices; it does not identify people. What speaker labels can and cannot tell you, the four ways they fail, and why the recording, not the software, decides how good they are.
Full verbatim keeps every um and false start; clean verbatim keeps every word of substance and nothing else. What each style includes, the same passage rendered both ways, and how to reach either from an automated draft.
The best open transcription model turns an hour of audio into text in about a minute and a half on one consumer GPU. Every number here was measured this month, with the corpus, the commit and the exact commands published.
Spanish transcribes at 3.7% word error and Hindi at 48.5%, on the exact engine FastScribe runs. Every number here comes from 200 FLEURS test clips per language, scored in the open, with the hard case, Hindi, told straight.
Convert subtitle files between SRT and VTT free, right in your browser. Your file never leaves your device. Paste or drop a file and download the converted version.
Check an SRT or VTT file for broken timestamps, overlaps, and ordering problems. Free, in-browser, with a line-by-line report. Nothing uploaded.
Check how long an audio or video file is and how big, right in your browser. See at a glance which FastScribe plan accepts it. Nothing is uploaded.
Work out how many hours of your own time a transcript will cost, typing it yourself against reviewing a software draft. Free, no signup, based on published ratios.
Count the words, characters and sentences in a transcript, with reading and speaking time. Handles caption files by stripping timings first. Nothing uploaded.
Getting a transcript
How to turn a recorded interview into an accurate, usable transcript, what affects quality, which route to take, and how to review the result properly.
How much time to budget for an hour of audio: 4 to 6 hours typing it yourself, minutes plus a review pass with software, days from a human service, and how to estimate your own recording.
Every genuinely free route to a transcript, manual, built-in tools, free tiers, and open-source models, with an honest account of what each one costs you in time, effort, and privacy.
A practical, start-to-finish guide to turning the meeting recording sitting in your downloads or cloud storage into an accurate written transcript.
Diarization separates voices; it does not identify people. What speaker labels can and cannot tell you, the four ways they fail, and why the recording, not the software, decides how good they are.
Choosing a tool
An honest comparison of every real way to get a free transcript in 2026: running an open speech model yourself, free tiers of transcription services, the tools already built into your phone and video platform, and doing it by hand.
Typing it yourself, running it through software, or paying a person each make sense in different situations, here is how to tell which one fits yours.
A real, working walkthrough of running OpenAI's Whisper speech recognition on your own machine, plus an honest look at when that beats paying for a transcription service.
The best open transcription model turns an hour of audio into text in about a minute and a half on one consumer GPU. Every number here was measured this month, with the corpus, the commit and the exact commands published.
Spanish transcribes at 3.7% word error and Hindi at 48.5%, on the exact engine FastScribe runs. Every number here comes from 200 FLEURS test clips per language, scored in the open, with the hard case, Hindi, told straight.
Comparing services
Otter is built around live meetings, and its file-upload allowances show it. Here is an honest look at where Otter fits, where it pinches, and which alternatives suit a folder of recordings better.
Rev sells professional transcription with a human guarantee, priced by the minute. For recurring everyday audio, that meter runs fast. Here is the honest arithmetic and the flat-rate alternatives.
TurboScribe is the closest thing FastScribe has to a direct rival, so this comparison is written carefully: where their deal is genuinely strong, where the fine print sits, and when an alternative fits better.
Recording quality and editing
The quality of a transcript is mostly decided before you ever upload the file, here is how to record so the words come back right.
Practical, step-by-step advice for salvaging noisy, echoey, or distant audio, before, during, and after transcription.
A practical, step-by-step workflow for turning a raw machine transcript into a clean, readable document, fixing names, cutting filler, labeling speakers, and exporting it in the right format to share.
Full verbatim keeps every um and false start; clean verbatim keeps every word of substance and nothing else. What each style includes, the same passage rendered both ways, and how to reach either from an automated draft.
Formats, timestamps, and privacy
A practical guide to the four most common transcript and subtitle export formats, what each one actually contains, where it works, and how to pick the right one before you download.
A practical guide to reading, choosing, and using time codes so a three-hour recording becomes something you can actually navigate.
A plain-language walkthrough of where your recordings go when you upload them to a transcription service, and how to judge whether that journey is acceptable to you.
Speech models invent fluent, correctly punctuated text where there was no speech. The four shapes it takes, where each one collects, and why the fix is a gate in front of the model rather than a better model.