What is the difference between Standard and Premium accuracy?

Fast, and accurate for most audio: a clear single speaker, a decent microphone, ordinary vocabulary. Higher accuracy, better punctuation and proper nouns. Worth it for interviews, names, accents and tougher recordings. The difference shows up on the hard parts of a recording — proper nouns, overlapping speech, accents and punctuation — rather than on the easy parts, so a clean solo recording often sounds identical on both.

Do I need a paid plan for Premium accuracy?

Yes. Available on Creator and above, from $20/mo. On the free tier the Premium option is visible but locked, and Standard remains available to everyone.

Should I tell it how many speakers there are?

If you know the number, yes. Declaring the count is the single most effective fix for a transcript where one person has been split across two labels or two people have been merged into one. If you genuinely do not know — an open meeting, a crowd — leave it on Auto-detect, because a wrong declared count is worse than no declaration.

Why is one speaker showing up as two?

Because separating voices is a judgement, and a change of tone, a moved microphone or a phone-quality patch in the middle of a recording all read as a different person. Declaring the exact speaker count is the fix; reassigning the crossed segments by hand afterwards is the fallback.

What do word timestamps actually give me?

A start and end time for every individual word rather than for each block of speech. That is what lets captions break where the speech breaks, lets you jump to the exact moment a word was said, and lets other software line the transcript up against the audio. Leave them on unless you have a specific reason not to.

Can I turn speaker separation off?

Yes, and you should for a recording with one voice — a dictation, a voicemail, a solo narration. There is nothing to separate, and separation applied to a single speaker is where the spurious second label comes from.

Transcription

Accuracy, speakers and timestamps

The three settings that decide whether a transcript is usable: when premium accuracy is worth paying for, how to get speaker labels you can trust, and what word timestamps unlock.

Lesson 2 · core · 7 min read

Open Transcription

Three settings do almost all the work here. One decides how well the words are heard, one decides whether the transcript knows who is talking, and one decides whether it knows exactly when. Everything else in the panel is a detail by comparison.

Accuracy

AudioTranscribe offers two accuracy tiers, and the honest way to choose is to judge the RECORDING rather than the importance of the job. A clear voice on a decent microphone comes back much the same either way; a noisy four-way call does not.

Tier: What it is for
Standard: Fast, and accurate for most audio: a clear single speaker, a decent microphone, ordinary vocabulary. Available on every plan.
Premium: Higher accuracy, better punctuation and proper nouns. Worth it for interviews, names, accents and tougher recordings. Available on Creator and above, from $20/mo.

The two tiers, and who can select them.

Write this

Not this

One person, quiet room, headset mic

Standard — the gap between the tiers has almost nothing to bite on here

Paying for Premium out of caution

Interview, two mics, some crosstalk

Premium — this is the case it exists for

Standard, then an hour repairing names

Conference call recorded through a laptop

Premium once, with the speaker count declared

Re-running the same tier hoping for a better result

A transcript that will be read, not published

Standard — good enough to search and skim is genuinely good enough

Premium on everything by default

One person, quiet room, headset mic

Write this:: Standard — the gap between the tiers has almost nothing to bite on here
Not this:: Paying for Premium out of caution

Interview, two mics, some crosstalk

Write this:: Premium — this is the case it exists for
Not this:: Standard, then an hour repairing names

Conference call recorded through a laptop

Write this:: Premium once, with the speaker count declared
Not this:: Re-running the same tier hoping for a better result

A transcript that will be read, not published

Write this:: Standard — good enough to search and skim is genuinely good enough
Not this:: Premium on everything by default

Note

Premium is a plan feature: Available on Creator and above, from $20/mo. On a free account the option is visible with a lock rather than hidden, so you can see what you are choosing between — and Standard is available to everyone, on every plan.

Speaker Diarization

Splits the transcript by who is talking and labels each segment with a speaker. On by default. Turn it off for a recording with one voice — there is nothing to separate, and a single speaker occasionally gets split in two. It is the setting people notice most, because a transcript that labels the wrong person is more annoying than a transcript that misspells a word.

Declares how many people are in the recording instead of letting it work that out. Only appears while diarization is on. Auto-detect, or an exact count from the picker — and if you know the number, say it. Separation is a judgement about how many distinct voices are present, and removing that judgement removes the two failure modes that come with it.

One person split into two labels: usually a change of tone, a moved microphone, or a stretch of worse-quality audio partway through.
Two people merged into one: usually similar voices, or one of them speaking only briefly.
Both are fixed by declaring the exact count — and neither is fixed by re-running with the same settings.

Tip

Turn separation OFF for a single-voice recording. There is nothing to split, and the only thing separation can do on a solo dictation is invent a second speaker.

Careful

Do not declare a count you are guessing at. Auto-detect being wrong is recoverable by editing a few segments; a confidently wrong count reshapes the whole transcript around a number that was never true.

Word Timestamps

Records a start and end time for every word rather than for each segment. This is what makes tight captions and word-level search possible later. They change nothing about the words you read on screen, which is why they are easy to dismiss — the difference appears in what you can do with the transcript afterwards.

Captions that break where the speech breaks rather than where a block of text happens to end — SubRip (.srt), WebVTT (.vtt) and JSON (.json) all carry timing.
Jumping to the exact moment a word was said, instead of to the top of the paragraph containing it.
Lining the transcript up against the audio in other software, which needs word-level times to do anything precise.

The next lesson is about what happens after all this: fixing what is left by hand, and choosing the export format that suits where the transcript is going.

Questions

What people ask about this

Free to start

Now go make one

Reading about a style description only gets you so far. The studio is free to use — write one sentence and hear what comes back.

Create a free account

1,000 credits every month. No card required.

All transcription lessons

Transcription

Accuracy, speakers and timestamps

The three settings that decide whether a transcript is usable: when premium accuracy is worth paying for, how to get speaker labels you can trust, and what word timestamps unlock.

Lesson 2 · core · 7 min read

Open Transcription

Accuracy

Tier: What it is for
Standard: Fast, and accurate for most audio: a clear single speaker, a decent microphone, ordinary vocabulary. Available on every plan.
Premium: Higher accuracy, better punctuation and proper nouns. Worth it for interviews, names, accents and tougher recordings. Available on Creator and above, from $20/mo.

The two tiers, and who can select them.

Write this

Not this

One person, quiet room, headset mic

Standard — the gap between the tiers has almost nothing to bite on here

Paying for Premium out of caution

Interview, two mics, some crosstalk

Premium — this is the case it exists for

Standard, then an hour repairing names

Conference call recorded through a laptop

Premium once, with the speaker count declared

Re-running the same tier hoping for a better result

A transcript that will be read, not published

Standard — good enough to search and skim is genuinely good enough

Premium on everything by default

One person, quiet room, headset mic

Write this:: Standard — the gap between the tiers has almost nothing to bite on here
Not this:: Paying for Premium out of caution

Interview, two mics, some crosstalk

Write this:: Premium — this is the case it exists for
Not this:: Standard, then an hour repairing names

Conference call recorded through a laptop

Write this:: Premium once, with the speaker count declared
Not this:: Re-running the same tier hoping for a better result

A transcript that will be read, not published

Write this:: Standard — good enough to search and skim is genuinely good enough
Not this:: Premium on everything by default

Note

Speaker Diarization

One person split into two labels: usually a change of tone, a moved microphone, or a stretch of worse-quality audio partway through.
Two people merged into one: usually similar voices, or one of them speaking only briefly.
Both are fixed by declaring the exact count — and neither is fixed by re-running with the same settings.

Tip

Turn separation OFF for a single-voice recording. There is nothing to split, and the only thing separation can do on a solo dictation is invent a second speaker.

Careful

Word Timestamps

Captions that break where the speech breaks rather than where a block of text happens to end — SubRip (.srt), WebVTT (.vtt) and JSON (.json) all carry timing.
Jumping to the exact moment a word was said, instead of to the top of the paragraph containing it.
Lining the transcript up against the audio in other software, which needs word-level times to do anything precise.

The next lesson is about what happens after all this: fixing what is left by hand, and choosing the export format that suits where the transcript is going.

Questions

What people ask about this

Free to start

Now go make one

Reading about a style description only gets you so far. The studio is free to use — write one sentence and hear what comes back.

Create a free account

1,000 credits every month. No card required.

All transcription lessons