Every transcription vendor quotes an accuracy figure and every figure was measured on someone else's audio.
This guide ranks the products on what actually varies between them: which languages and accents hold up, how well two people talking over each other are separated, where the recording is processed, and whether the vendor may train on what you upload.
Vendors can pay for visibility on this page. It never changes what an entry
says about a product, including the criticism, and we earn nothing when you click through to a
vendor. How that works.
In short
What transcription software does
Transcription software converts recorded speech into text, labels who spoke when, and returns a timed, searchable document that you can correct, caption, translate or export elsewhere.
Five things, in this order. Feature counts are not among them: they are the least useful
comparison in software, because every vendor ticks every box.
01
Setup effort in transcription software
What the first ninety days of a transcription software rollout cost in hours, not in licence fees. A product that needs a partner engagement before it does anything is a different purchase from one a team configures in an afternoon.
02
What transcription software really costs
What the bill becomes once the modules a normal buyer of transcription software needs are added, and whether you can read that number without a sales conversation.
03
Getting your data out of transcription software
How your own data comes back out, in what format, and whether that export is included in the transcription software contract or billed as a project.
04
Independence from the vendor
Whether you can buy transcription software, run it and leave it on your own terms. This test decides most of the order on this page, and it is why the largest vendors in transcription software often sit below the smaller ones.
05
Who the product is built for
The size and shape of company each transcription software product was actually built for. Most regret in software comes from buying for a company you are not yet.
The fourth test decides most of the order on this page, and it is the reason the largest
transcription software vendors sit below the smaller ones. A product with a published price, an export
that works and no mandatory implementation partner is a product you can leave.
A platform suite that arrives with a quote, a partner and a two-year commitment may well be
the better software and is still the harder decision to reverse. We rank transcription software for the
buyer who has to live with that decision without a procurement department, which is a stated
bias rather than a hidden one.
We do not publish a score out of ten. A number like 8.4 is a judgement dressed as a
measurement, and nobody can check it.
What you can check is on this page: what each transcription tool costs, where the vendor is
established, whether the price is published, and what we think it is bad at. Our full method
is on the how we work page.
Pay per minute of audio, published; self-hosted deployment quoted
—
Engineering teams processing large volumes of audio, or needing on-premise
Office users who want to upload a recording and read it
Country is where the vendor is headquartered or contracts from, which is a
different question from where your data is hosted. Where the two tell different stories, the
entry says so.
Dutch transcription and subtitling with processing kept in Europe
Ranked #1 of 12 in Best Transcription Software in 2026.
Published pricingEurope
Amberscript keeps processing and storage in the EU and treats Dutch, German, French and the Nordic languages as first-class rather than as a long tail. When automated output is not sufficient, the same file can be sent to a human transcriber under the same contract, which matters for court, research and broadcast work.
The pricing model punishes high-volume everyday use, so pair it with something cheaper for internal meetings and keep this for the recordings that count.
What stands out
EU processing
Human option
Subtitles
Where it costs you
Per-minute cost adds up fast for daily meeting transcription
The editor is functional rather than pleasant
Right for
European organisations transcribing regulated or multilingual recordings under European rules
Wrong for
Teams wanting unlimited transcription for a flat seat price
NetherlandsPay per minute of audio, published; human transcription priced separately
Transcription and subtitles priced per minute with a usable editor
Ranked #2 of 12 in Best Transcription Software in 2026.
Published pricingEurope
Happy Scribe is straightforward to buy and to leave: per-minute rates on the site, a subtitle editor that handles the formats broadcasters ask for, and both machine and human tiers. It suits one or two clear voices.
Put four people around a table and the speaker labels start drifting, which costs more correction time than the price saved. Test your own worst recording during the trial rather than a clean one.
What stands out
Published pricing
Subtitle formats
EU vendor
Where it costs you
Speaker separation struggles on recordings with several voices
Quality varies widely across the languages listed
Right for
Podcasters, researchers and video teams buying transcription by the minute
Wrong for
Multi-speaker meetings where the speaker labelling must be reliable
IrelandPer minute of audio or a monthly plan, published
Ranked #3 of 12 in Best Transcription Software in 2026.
Self-hostablePublished pricingEurope
Speechmatics is an engine, and the two reasons to choose it are accent coverage and deployment. It handles varieties of English that models trained on American broadcast speech routinely fail, which shows up immediately on regional UK, Irish, Indian or African English.
Running it in your own container removes the residency argument entirely. Everything a user touches, you build, so this is a decision for an engineering team with a roadmap, not a purchase for an office.
What stands out
On-premise option
Accent coverage
API
Where it costs you
No editor, library or interface for non-technical users
Container deployment is priced and negotiated separately
Right for
Product and platform teams needing recognition inside their own systems
Wrong for
A team that just wants to upload a file and read it
United KingdomPer hour of audio, published; self-hosted container quoted
French speech API with European processing and per-hour pricing
Ranked #4 of 12 in Best Transcription Software in 2026.
Free tierPublished pricingEurope
Gladia offers a speech API with European processing, published per-hour pricing and a free tier you can benchmark on before talking to anyone. Diarisation, translation and real-time streaming are all covered.
Set against Deepgram and Speechmatics it is the smaller company: fewer deployment options, less operational history, and a smaller support organisation. For an EU product team that needs the processing location settled without an enterprise negotiation, that trade is often the right one.
What stands out
EU processing
Developer API
Real-time
Where it costs you
Younger company with thinner tooling around the API
Nothing here for a non-technical end user
Right for
European product teams embedding transcription with processing kept in Europe
Wrong for
Individuals or offices transcribing their own occasional recordings
FrancePer hour of audio, published; free tier for testing
Records and transcribes meetings, then clips the useful minutes
Ranked #5 of 12 in Best Transcription Software in 2026.
Free tierPublished pricingEurope
tl;dv joins your calls, records and transcribes them, and its useful trick is clipping a moment and sending it to a colleague instead of asking them to watch an hour.
The free tier is generous enough to test properly and the company is in Berlin, which shortens the data conversation for European buyers. It is scoped to meetings only. And a bot with a name in the participant list still needs everyone's agreement to record.
What stands out
Meeting recorder
Free tier
Berlin vendor
Where it costs you
Meetings only; uploaded audio and archives are out of scope
A recording bot in the call makes consent your responsibility
Right for
Sales and product teams reviewing recorded calls together with their colleagues
Wrong for
Interviews, field recordings or any audio outside a meeting
Editorial transcription built for newsrooms working to deadline
Ranked #6 of 12 in Best Transcription Software in 2026.
Published pricingEurope
Trint understands the editorial job: highlight as you listen, keep the audio anchored to every word so a quote can be verified, and hand the story to a colleague without emailing files. Newsrooms stay with it for that.
The seat-based model is the problem for everyone else, because a licence bought for a busy month keeps billing during a quiet one, and per-minute vendors like Sonix or Happy Scribe will be cheaper for irregular use.
What stands out
Newsroom workflow
Story editor
Collaboration
Where it costs you
Per-seat pricing is expensive for occasional transcription
Overspecified for anyone outside editorial work
Right for
Newsrooms and content teams turning recorded interviews into publishable copy quickly
Wrong for
Organisations transcribing only a few recordings each month
United KingdomPer seat per month, published; enterprise quoted
Pay per hour with the best correction editor in the category
Ranked #7 of 12 in Best Transcription Software in 2026.
Published pricingNorth America
Correcting a machine transcript is where the hours go, and Sonix has spent its effort exactly there: word-level audio sync, fast keyboard navigation, find-and-replace across speakers, and clean exports. Buying by the hour with no subscription suits irregular volumes.
The limitation is jurisdictional rather than functional. If your recordings contain health data, legal privilege or anything a regulator will ask about, the processing location is the first question and this is the wrong answer.
What stands out
Per-hour pricing
Strong editor
Many formats
Where it costs you
US company and US processing, with no EU residency guarantee
Accuracy on heavily accented audio is ordinary
Right for
Anyone who corrects transcripts by hand and values the editing time
Wrong for
Regulated European content that must stay inside the EU
United StatesPay per hour of audio, published; optional per-seat subscription
Edit the audio by editing the transcript, then publish it
Ranked #8 of 12 in Best Transcription Software in 2026.
Published pricingNorth America
Descript's premise is that editing text is easier than editing a waveform, and for producing episodes it works. Removing filler words, cutting a tangent and regenerating a mis-spoken line all happen in the transcript. Judge it as an editing suite.
As a transcription service the accuracy is mid-table, the included hours are modest, and the synthetic voice features introduce consent and disclosure questions that a compliance team will want to discuss before anyone uses them.
What stands out
Audio editing
Podcast workflow
Voice cloning
Where it costs you
Transcription accuracy is a means to an end, not the product
Included hours run out quickly on a per-seat plan
Right for
Podcast and video teams editing published audio through the text
Wrong for
Anyone needing an accurate record of what was said
United StatesPer seat per month with included transcription hours, published
Cheap meeting notes with a free tier most teams start on
Ranked #9 of 12 in Best Transcription Software in 2026.
Free tierPublished pricingNorth America
Otter.ai wins on friction: connect a calendar, and meetings are recorded, transcribed and summarised without anyone doing anything. At the price, for English, that is hard to beat. The two cautions are real.
Non-English audio produces transcripts that cost more to fix than to redo, and the terms governing how recordings may be used to improve the service deserve reading before client conversations, HR meetings or anything under privilege go into it.
What stands out
Meeting notes
Free tier
Calendar sync
Where it costs you
English-first, with weak performance in other languages
Read the terms covering what recordings may be used for
Right for
English-speaking teams wanting cheap automatic notes from every internal meeting
Wrong for
Confidential or multilingual recordings in a regulated organisation
United StatesFree tier; per seat per month, published
Human transcription per minute when the machine pass is not enough
Ranked #10 of 12 in Best Transcription Software in 2026.
Published pricingNorth America
When accuracy has to survive being quoted, a human transcript is still the answer, and Rev has the scale to deliver one overnight at a published rate. That is a genuine capability and few competitors match the turnaround.
The confidentiality terms are the thing to examine, since a person somewhere reads the file. Processing is American, non-English work is thin, and the automated tier is competent without being a reason to choose them.
What stands out
Human transcripts
Per-minute rates
Fast turnaround
Where it costs you
Human transcription means people read your recordings
Non-English coverage is limited and American processing throughout
Right for
Legal, academic and media work where a machine transcript is insufficient
Wrong for
European recordings that may not leave the jurisdiction
United StatesPer minute of audio, published; separate human and automated rates
Captioning and transcription for universities and courts, with humans
Ranked #11 of 12 in Best Transcription Software in 2026.
Pricing on requestMiddle East
Verbit sells to institutions that must caption everything and be able to prove it: lecture capture, legal proceedings, service levels and an audit trail, with human review built into the workflow rather than sold as an extra.
That is a defensible purchase when accessibility is a legal duty. It is inaccessible commercially to anyone smaller, since pricing is negotiated against committed volume, and the processing chain needs proper legal review in a European context.
What stands out
Accessibility
Human review
Enterprise
Where it costs you
Quoted pricing with committed volumes and no self-serve route
Processing arrangements span several jurisdictions
Right for
Universities and courts with a statutory accessibility obligation to meet
Wrong for
Small teams wanting to buy transcription with a card
IsraelQuoted per organisation, per minute of media
Fast speech API with a self-hosted option for regulated buyers
Ranked #12 of 12 in Best Transcription Software in 2026.
Self-hostablePublished pricingNorth America
Deepgram is built for throughput and low latency, which is why it turns up behind voice agents and call analytics rather than in front of end users.
Cloud pricing is published per minute and the self-hosted deployment is the reason regulated buyers shortlist it, since audio can stay inside the bank or the hospital. Test your own languages properly. The published benchmarks are English-heavy and the gap on other languages is wider than the marketing implies.
What stands out
Low latency
Self-hosted option
Developer API
Where it costs you
An API only; there is no application for end users
Non-English accuracy varies more than the benchmarks suggest
Right for
Engineering teams processing large volumes of audio, or needing on-premise
Wrong for
Office users who want to upload a recording and read it
United StatesPay per minute of audio, published; self-hosted deployment quoted
Transcription software converts recorded speech into text, labels who spoke when, and returns a timed, searchable document that you can correct, caption, translate or export elsewhere. The differences that matter are rarely in the feature list, so this is
the order we would work through them.
01
Decide whether you need a published price
11 of the 12 tools here publish what they cost; the other 1 quote per organisation, which means a sales conversation before you can compare anything. If you are buying without a procurement function, start with the ones that publish: Amberscript, Happy Scribe, Speechmatics, Gladia, tl;dv, Trint, Sonix, Descript, Otter.ai, Rev, Deepgram.
02
Work out what the first ninety days cost in time
Licence cost is the number in the contract; setup effort is the number that surprises people. Ask every shortlisted vendor who does the configuration, how long it took the last customer of your size, and what happens if that person leaves halfway.
03
Check the exit before the entry
Ask for an export of your own data in a format you can open, and ask whether it is included or billed as a project. A vendor that hesitates here is telling you what renewal negotiations will feel like in three years.
04
Match the tool to the size you are, not the size you plan to be
Most regret in this category comes from buying for a headcount that never arrived. The entry-level products here are not worse; they are aimed at a different company.
05
Decide how much the jurisdiction matters
These 12 vendors are established in 7 countries across 3 regions (Europe 6, North America 5, Middle East 1). Where a vendor is established decides which government can compel access to what it holds, which is a different question from where the servers are. For most buyers that is a factor, not a veto.
Accuracy claims are marketing until you measure them on your own audio
Word error rate is quoted the way fuel consumption used to be: measured under conditions you will never reproduce. Clean studio speech, one speaker, a standard accent, no jargon. Your audio is a meeting room with a fan, four people, two of them non-native speakers, and forty product names the model has never seen. Benchmark it yourself.
Take three recordings that represent your worst case, run them through every shortlisted vendor's trial, and count the corrections needed to make each transcript usable. Amberscript, Happy Scribe and Sonix all let you do this for the price of a few files. The differences you find will be larger than the differences in the published figures, and they will point the other way surprisingly often.
Test your three worst recordings, never the vendor's sample file.
Count minutes spent correcting, not the accuracy percentage claimed.
Include your own product names and jargon in the test and see what survives.
Language and accent coverage is where the list on the pricing page lies
A supported-languages list counts languages the model will attempt, not languages it does well. The gap is largest exactly where European buyers need it: Dutch with English loanwords, Swiss German, Flemish, regional Italian, and any conversation where speakers switch language mid-sentence, which no vendor handles gracefully. Accent is the same problem inside one language.
Speechmatics built its reputation on varieties of English that models trained on American speech mishandle, and Amberscript on European languages that American vendors treat as an afterthought. Otter.ai is very good in English and much weaker outside it. Ask for the error rate in your specific languages rather than the headline, and if the vendor cannot produce one, that is itself the answer.
Ask for accuracy figures per language, not one global number.
Test code-switching if your meetings mix two languages, because most tools fail it.
Check whether speaker labels stay stable when an accent is unfamiliar.
Where the audio is processed decides everything else for regulated buyers
A recording of a patient consultation, a disciplinary hearing or a privileged legal call is not a file you may upload anywhere. For those buyers the processing location is not one criterion among several, it is the whole decision, and it eliminates most of this page. Three shapes work. A European vendor processing in Europe, such as Amberscript or Gladia.
An engine you run inside your own network, which Speechmatics and Deepgram both offer as a container. Or a human service under a contract your legal team has actually read, as with Verbit. Ask where the audio is processed, where it is stored afterwards, how long it is kept, and which sub-processors touch it, then get all four answers in the contract rather than in an email.
Ask for the processing region and the storage region separately; they differ.
Get the full sub-processor list, including any human transcription supplier.
Set a retention period and verify that deletion actually removes the audio.
Whether the vendor trains on your recordings, and who consented to being recorded
Two questions that rarely appear in a comparison table. First, what the terms let the vendor do with your audio: many services reserve the right to use recordings to improve their models, sometimes on by default, sometimes off only on a paid plan. Otter.ai's terms are worth reading in full before client calls go into it. Second, consent.
A recording bot that joins a call, as tl;dv and every meeting recorder does, records everyone present, and under GDPR the obligation to inform and to have a lawful basis sits with you, not with the vendor. Neither question is answered by the accuracy score, and both are the ones that surface after an incident rather than before a purchase.
Find the clause on model training and get any opt-out confirmed in writing.
Decide who announces the recording, and check the bot is visible to all participants.
Write a retention rule for recordings and transcripts, then enforce it automatically.
What goes wrong most often when buying transcription software
Choosing on the published accuracy figure. It was measured on clean speech in one accent and tells you nothing about your meeting room.
Assuming an EU data centre means EU processing. The inference may happen elsewhere and the sub-processor list is where you find out.
Buying seats when your volume is irregular. Per-minute vendors cost less for anyone who transcribes in bursts rather than daily.
Letting a recording bot into external calls without telling the other participants. The consent obligation is the customer's, and no vendor absorbs it.
07
Frequently asked questions
10 answers
What is the best transcription in 2026?
Amberscript leads our ranking of 12. Handles the European languages other vendors treat as an afterthought, and will put a human on the file when the automated pass is not good enough for a court or a broadcaster.
Processing and storage stay in the EU, which is the reason most of its customers are there. The editor is plain, and per-minute pricing gets expensive for someone transcribing every meeting.
How did you rank these transcription tools?
On what separates products after the demo: how much setup the first ninety days take, what the price becomes once the modules a normal buyer needs are added, how your data comes back out, whether you can buy and leave it without a partner engagement, and who the product is genuinely for.
That fourth test is why the large platform suites usually sit lower here than their market share would suggest. Not on feature counts, and not on a score we invented.
Which transcription tools publish their pricing?
11 of the 12, with the pricing model each one publishes:
Amberscript: Pay per minute of audio, published; human transcription priced separately.
Happy Scribe: Per minute of audio or a monthly plan, published.
Speechmatics: Per hour of audio, published; self-hosted container quoted.
Gladia: Per hour of audio, published; free tier for testing.
tl;dv: Free tier; per seat per month, published.
Trint: Per seat per month, published; enterprise quoted.
Sonix: Pay per hour of audio, published; optional per-seat subscription.
Descript: Per seat per month with included transcription hours, published.
Otter.ai: Free tier; per seat per month, published.
Rev: Per minute of audio, published; separate human and automated rates.
Deepgram: Pay per minute of audio, published; self-hosted deployment quoted.
The other 1 quote per organisation.
Is there a free transcription tool?
Gladia, tl;dv, Otter.ai offer a free tier or a free self-hosted edition. Read what the free tier excludes before you plan around it.
Which transcription tools can you host yourself?
Speechmatics, Deepgram. The other 10 are sold as a hosted service only, which means the question of where your data sits is answered by the vendor, not by you.
Where are these transcription vendors established?
In 7 countries across 3 regions: Europe 6, North America 5, Middle East 1.
Amberscript is established in the Netherlands.
Happy Scribe is established in Ireland.
Speechmatics is established in the United Kingdom.
Gladia is established in France.
tl;dv is established in Germany.
Trint is established in the United Kingdom.
Sonix is established in the United States.
Descript is established in the United States.
Otter.ai is established in the United States.
Rev is established in the United States.
Verbit is established in Israel.
Deepgram is established in the United States.
Establishment decides whose courts and whose disclosure laws apply, which is a separate question from where the data is hosted.
What should you use instead of Amberscript?
Happy Scribe and Speechmatics are the next two on this page.
Happy Scribe is for Podcasters, researchers and video teams buying transcription by the minute; Speechmatics is for Product and platform teams needing recognition inside their own systems. All 12 are ranked here with what each one is bad at.
Who should not buy Amberscript?
Teams wanting unlimited transcription for a flat seat price. Per-minute cost adds up fast for daily meeting transcription.
Do you get paid for these rankings?
Vendors can pay for visibility, which affects where and how prominently a product appears. It does not change a word of what the entry says about that product, including the criticism, and it cannot buy inclusion for something that does not belong in the category.
We take no commission when you click through to a vendor and we do not know whether you bought anything. The full arrangement is on our disclosure page.
How often is this transcription guide updated?
Whenever the facts move: a price change, an acquisition, a product that stops being maintained. The published and updated dates at the top of the page are real, and a review means someone went back to the vendor documentation rather than bumping a date.
These 12 products are the ones we judged worth ranking in transcription. If yours belongs here and is missing, tell us what it does and who it is for, and we will look at it. Inclusion is an editorial call and it is not for sale — but nobody gets considered for a list they were never put in front of.
People land on this page with a shortlist to make, not a browsing habit to feed. That is a narrower audience than a banner reaches and a far more decided one.
Written by us, about you
We describe the product in our own words, say who it suits and say who it does not. A vendor never writes the entry and never sees it before it goes up.
A correction costs nothing
If a fact about your product is wrong here, tell us and we fix it, whether or not there is any money between us. That offer is older than any commercial arrangement on this site.
Placement is separate, and disclosed
Where a product sits in the ranking can be paid for, and the notice above the list says so on every page. What the entry says about the product is not for sale at any price.
We use analytics cookies only if you agree. See our privacy policy.
