Pronunciation Checker Tools | 9 Compared, 6 Free, 5 That Score You

Search for a free English pronunciation checker and you get list after list of tools. The trouble is that the things on those lists are not the same kind of tool. Some score the voice you record. Some only play a native speaker saying the word. Some just turn text you type into audio. They all end up in the same table with nothing separating them.

The conclusion first. Of the 9 tools below, 5 put a number on your own voice. The rest are for hearing a model, making a model, or turning what you said into text, and none of them score anything. That is not a criticism. They have different jobs.

There is a second thing that gets mixed into the free shelf: apps that are free for 7 days and paid after that. Of the 9, 6 have a tier you can keep using with no deadline, There is a second thing that gets mixed into the free shelf: apps that are free for 7 days and paid after that. Of the 9, 6 have a tier you can keep using with no deadline, 2 become paid once the trial ends, and 1 does not publish its prices at all., and 1 does not publish its prices at all.

Quick recommendations by purpose: If you want to see how you actually pronounced words in a real lesson rather than in a prepared sentence, use SpeakCoach. If you want a single measurement right now, use AnyToSpeech, which requires no account or installation. If you want something to use every day, try ELSA Speak. If you only want to hear the correct pronunciation of a word, use Cambridge Dictionary. The reasons are in the table and the sections below. If you want to see how you actually pronounced things inside a real lesson rather than inside a prepared sentence, SpeakCoach. If you want one measurement right now, AnyToSpeech, which needs no account and no install. If you want something to keep doing every day, ELSA Speak. If you only want the correct sound of a word, Cambridge Dictionary. The reasons are in the table and the sections below.

Key takeaways

• Tools called pronunciation checkers come in 4 kinds: ones that score you, ones that play a model, ones that make a model, and ones that turn your speech into text
• Of the 9 here, only 5 put a number on your own voice. The others score nothing
• 6 have a free tier with no deadline; 2 become paid once the trial ends
• 1 of them does not publish its prices
• If all you want is the correct sound of a word, you do not need a scoring tool. A dictionary is enough
• A score shows you where you are off right now. Raising the number is not the goal in itself

Four different kinds of tools are all called pronunciation checkers. all get called a pronunciation checker

The tools that come up in search split into 4 groups by what they actually do. Split them first and you can decide for yourself which one to open.

• Scores your voice — analyses what you said and returns a number or a mark showing where you were off
• Plays a model — lets you hear how a native speaker actually says it. No score
• Makes a model — reads out English you type and turns it into audio. No score
• Recognises only — turns what you said into text. It never says whether you were right

This split is not something invented for the sake of an argument. One text-to-speech service states in its own article that it has no automatic pronunciation scoring and no way of flagging where you went wrong, and it still ranks high when people search for a pronunciation checker. The tool is being honest. The roundups are the ones mixing things together.

Once you can see there are 4 kinds, the order follows. Hear a model so you know the correct sound. Get scored on whether you can produce it. When something needs fixing, go back to the model. Scoring on its own goes nowhere, because if you do not know the correct sound Once you can see that there are four kinds, the order becomes clear. Hear a model so you know the correct sound. Get scored on whether you can produce it. When something needs fixing, go back to the model. Scoring on its own goes nowhere, because if you do not know the correct sound, there is nothing to aim for..

Free and free trial are not the same shelf

Some of what gets listed as free is a trial with an expiry date on it. The table below uses two different labels for these: has a free tier, and paid after a trial.

BoldVoice, for example, says on its own site that the first 7 days are free, and Speak has a 7-day free trial after which it splits into Premium and Premium Plus. Both are good apps. If what you are looking for is something you can keep using for nothing, they belong on a different shelf.

One more thing, about price itself. Inside ELSA Speak's own domain, 3 different prices for the same subscription are visible at the same time. Some campaign is always running, so which one you see depends on when you open the page. Read the figure off the checkout screen on the day, not off a roundup written earlier.

If you want to use Google without installing anything, use its web-based tools.

Sometimes you only want to check a sound and you do not want another app on your phone. Google will play a word for you straight from the search box, and there is also a feature that listens to you say the word and tells you how you did. That second one is the part to watch. Google itself describes it as a limited beta experience, and it runs on language pairs, where you practise a language other than the one you already speak. There is no pairing in which an English speaker practises English, so that feature is not something you can open.

What does work without installing anything is the plain speech-to-text side. Open Google Translate, set the input language to English, say the word, and read back what it typed. That tells you whether the machine heard the word you meant, and nothing beyond that. It is a pass or a fail, not a score: it will not tell you which syllable you pressed on, and it will not tell you which sound was off. For anything finer, you need one of the tools that actually scores you.

How to choose a pronunciation checker

Settling your own conditions first is faster than comparing 9 tools. There are 4 things to decide.

• Free forever, or is paid acceptable — the free shelf has 7-day trials mixed into it, so look for an expiry date first
• What you want measured — a word, a fixed sentence, or a conversation. Tools differ in the unit they can handle
• How fine you need it — is a number enough, or do you need to know which sound was off? Some go down to the individual sound; others stop at the word
• Whether the recording is kept — if you want to listen back later, pick one that keeps it; if you do not want your voice stored, pick one that states it does not

The fourth one gets overlooked. What happens to your recorded voice is often not written anywhere on the official pages. In the table, that column is marked ※.

The 9 at a glance

The 5 ways to assess your own voice

What changes when a score comes back is that you can see the gaps you cannot hear in yourself. While you are speaking, you have no access to how you sound to the person listening. The tools here use a machine to point out that difference for you.

SpeakCoach

★What makes it different is that it measures how you pronounced things inside a lesson you actually took, not inside a practice sentence. The recording stays, so you can listen back later

Open it Open it when you want to see how you spoke in a real lesson rather than in a prepared sentence. It covers the sound you make while you are still working out what to say, which reading practice never reaches. to see how you spoke in a real lesson rather than in a prepared sentence. It covers the sound you make while you are still working out what to say, which reading practice never reaches.

SpeakCoach official page

ELSA Speak

A learning app with scoring built in. The official site talks about a learning path optimised for each person and immediate personalised feedback in every lesson. There is a free tier

★3 different prices for the same thing appear inside the official domain at the same time. A campaign is always running, so which price you see depends on when you open it

Open it Open it when you want a daily lesson rather than a one-off measurement. Because a learning path is laid out for you, you do not have to decide what to work on., rather than a one-off measurement. Because a learning path is laid out for you, you do not have to decide what to work on.

the official ELSA Speak page

AnyToSpeech Pronunciation Test

★Scores on 4 dimensions: completeness, accuracy, fluency and pronunciation quality. Read out the sentence it shows you and it returns a score out of 100 plus word-by-word feedback, and it detects errors down to the individual sound. No sign-up, and it works the moment you open it

Price: the pronunciation test itself is free and needs no account or registration; the official answer is that it is completely free. A premium tier that removes the daily limit starts at $7 a month

★It is one of the few tools that states outright that it does not keep recordings. The official wording is that audio is sent for analysis and deleted immediately, and that voice recordings are never stored. The interface is available in 12 languages. The official wording is that audio is sent for analysis and deleted immediately, and that voice recordings are never stored. The interface is available in 12 languages

Open it when — you want one measurement right now. No account, no install: open the page, read the sentence it shows you, done. It’s useful for finding out where you currently stand.

The official AnyToSpeech Pronunciation Test page

BoldVoice

★One of the few tools whose official site names the things it scores: pronunciation, stress and clarity. It records you saying phrases and having conversations, then analyses them.

Price: the first 7 days are free. No figure appears on the front page

Open it Open it when you want to be asked to repeat yourself less often at work. Because the official site names what it scores, the numbers map onto something you can act on. to be asked to repeat yourself less often at work. Because the official site names what it scores, the numbers map onto something you can act on.

BoldVoice's official page

Speak

The official site says it returns not only pronunciation but how natural your phrasing sounds. Premium Plus generates lessons out of your own mistakes

Price: a 7-day free trial, then 2 tiers, Premium and Premium Plus. The amounts are not published

Open it Open it when you do not want to focus solely on pronunciation but also want to see whether your phrasing is natural. Feedback comes back inside a spoken exchange, which feels different from drilling words. but to see whether your phrasing is natural at the same time. Feedback comes back inside a spoken exchange, which feels different from drilling words.

Speak official page

Tools for hearing a model, and tools for making one

Scoring tools on their own are not enough. Being told that a sound is off is no help if you do not know what the correct sound is. So here are the tools for hearing a model, and for making one.

Cambridge Dictionary

★A dictionary that plays British and American English side by side, with the phonetic transcription next to them. The official site says the audio is not synthesised but recorded by real people, and that and that it covers every word from CEFR A1 to C2

Open it Open it when you want to check the correct pronunciation of a word.. The transcription is printed alongside, so you can check the sequence of sounds with your eyes as well as your ears.

Cambridge Dictionary official page

YouGlish

★Instead of a dictionary recording, it lines up real video clips of the word being said. You can filter by part of speech, by speaker gender and by person (Obama or Clinton, for example), so you can hear how it shifts with context

★The official About page says nothing at all about price and asks for donations instead. There is no interface language switcher on the About page or on the front page either (English is one of the languages you can search in)

Open it Open it when the dictionary recording does not settle it for you.. The same word changes with the speaker and the situation, so lining up several real instances gives you a sharper outline than one dictionary recording does.

YouGlish's official page

NaturalReader

★A tool that reads English you type and makes a model out of it. It does not score. Long documents and PDFs can be turned into audio as they are, so you can supply your own practice material. It runs in the browser, as a phone app, and as a Chrome extension

Price: Price: The official site states that it offers a free-forever plan and that no credit card is required. Paid plans are divided by use case (Personal, Commercial, and EDU).. Paid plans are split by use (Personal, Commercial, EDU)

Open it Open it when you want to supply your own material. to supply your own material. Coursebook English or a document you have to read at work: paste it in and it becomes audio. Playing it alternately with your own recording is an easy way to compare the two.

NaturalReader official page

The tool only recognises English.

Google Translate

★The official site only ever describes speech as an input method: The official site describes speech only as an input method—for example, by saying that users can type, speak, or write by hand, or use voice input. Nothing on the page mentions evaluating pronunciation.. Nothing on the page mentions evaluating pronunciation

★Recognised correctly does not mean pronounced correctly. Getting through is a pass or a fail, not a score

Open it Open it when you want a rough sense of whether you are getting through, using what is already on your device. a rough sense of whether you are getting through, using what is already on your device. It suits a very coarse check before you commit to a dedicated tool.

the official Google Translate page

Whether your speech turned into the right text is not a score. Recognition succeeding only means the machine could make it out. No number comes back, and neither does any indication of what was off. It works for a rough check on whether you get through, but assume it gives you nothing you can act on.

Why your first language keeps pulling the sound back, and how to fix rhythm

The first wall you hit when you try to fix your pronunciation is that you cannot hear that you are off. When English has a sound your first language does not, you learn it as the nearest sound your first language does have, and that substitute settles in as the sound of the word. Inside your own head it is correct, so there is nothing to be suspicious of until somebody tells you.

In second language research, this noticing step has long been treated as the entry point for learning. Exposure on its own does not install a form; it is only when attention lands on the form that it gets taken in.

This is where a scoring tool earns its place. More than the number itself, what works is finding out that a sound you were sure you had right is off. Put the other way round: if you only let models wash over you, no noticing happens, and the sound stays exactly as it was.

Rhythm and where the stress falls

Individual consonants and vowels are the part people worry about, but what mostly decides whether you get through is Individual consonants and vowels are what people worry about, but what mostly decides whether you get through is which syllable you stress and the rhythm of the sentence as a whole. and the rhythm of the sentence as a whole. Put the stress in the wrong place and a word can be heard as a different word, even when every sound inside it is correct.

Among the tools listed here, the ones that score on several dimensions look at fluency and stress as well as at the accuracy of the sounds. Scoring high on single words and still not getting through in conversation happens because the unit being measured is a different one.

And you do not have to lose your accent. Most of the people who speak English are not native speakers, so it is simply not true that you have to sound like one to be understood. Getting to the point where the other person does not have to ask you to repeat yourself is enough.

What AI scoring is actually looking at, and where accent fits

Several of the scoring tools rest on the same design. One widely used engine is the pronunciation assessment in Azure AI Speech, and because its specification is public you can read straight off what it looks at. When a score comes back lower than you expected, this breakdown tells you where to look.

4 scores come back

The point is that these 4 look at different things. Every individual sound can be accurate and fluency will still drop if your pauses are unnatural. You can speak smoothly and accuracy will still drop if individual sounds have been substituted. There is no single ruler called good or bad pronunciation.

Accuracy goes all the way down to the individual sound. For hello, for instance, the expected sounds are h ɛ l oʊ, and the result can come back saying that what you produced in the second position was ə rather than ɛ.

The overall score is decided by your lowest score

This is the most misunderstood part. The overall score is not the average of the 4. The 4 are sorted from lowest to highest and then weighted differently, starting from the lowest.

So 3 scores of 90 and 1 of 50 pull the overall figure sharply toward 50. When the number will not move, the fast route is not to raise the average but to identify the single lowest one and fix only that.

Per word, the type of error comes back as well. A word scoring under 60 on accuracy is flagged as mispronounced, and the other types are omission, insertion, unexpected break, missing break and monotone. Omission and insertion only appear when you hand the tool the text you are going to read. You can identify omissions or insertions by comparing the text you were given with what you actually said.

Accent detection is a different thing from scoring

There is a well-known party trick where an AI guesses your country from your English accent. That is not pronunciation scoring. It classifies which region your speech resembles, and it says nothing about better or worse.

Pronunciation assessment measures distance from an expected sound pattern. Having an accent is not itself penalised; the relationship is that moving away from the expected sounds lowers accuracy. Erasing an accent and raising a score are not the same project. If the goal is to know which country you sound like, one visit to a detector site settles it. If the goal is to be understood, the breakdown of the score is the thing to look at.

A score comes back. A fix does not

Use a scoring tool and you almost always get a number. But why the number dropped, and what to do differently with your mouth tomorrow, are separate questions. Tools that go down to the individual sound will tell you which sound was off; how to produce that sound is something you go and find out for yourself.

And a sound that has already set does not come loose just from hearing the correct one. The idea that learners build a system of their own that stays put, and hardens if left alone, has been argued for half a century.

An analysis pooling classroom corrective feedback reports that pushing learners to reformulate for themselves has a larger effect than supplying the correct form for them to hear. ★What that analysis compared was ways a teacher gives feedback, not machine scoring. What follows from it is only that having a chance to say it again yourself, after noticing, makes it stick. It says nothing about the effect of a score.

Also, scores rise when you read the same sentence over and over. Whether what improved is your pronunciation or just your familiarity with that one sentence is not something the score can separate. 90 on the third attempt does not mean 90 on a sentence you have never seen.

Combining it with shadowing: How it improves your listening skills

Every tool so far measures a bounded piece of sound. One word, one sentence, a fixed length read out and scored. That is measuring at a point. Real conversation, on the other hand, does not stop. Words that are right on their own can collapse once they are joined up.

Shadowing fills that gap. You listen to a model and say the same thing aloud a beat behind it, copying the rhythm as a whole instead of fixing sounds one at a time. Where sounds join and become another sound, where they disappear, where the pressure falls: all of it goes in together.

Paired with a scoring tool, the roles divide. Shadowing builds the flow; scoring finds Paired with a scoring tool, the two methods serve different roles. Shadowing builds the flow; scoring identifies which sound differs from the reference; then you go back to the model for that sound alone. With only one of the two, you drift to one side: good flow with individual sounds still off, or accurate sounds that come out stiff in conversation.; then you go back to the model for that sound alone. With only one of the two you drift to one side: good flow with individual sounds still off, or accurate sounds that come out stiff in conversation.

You can supply the material yourself with a tool that makes models. Coursebook English, or the phrasing you actually use at work: have it read aloud and it becomes shadowing material. Working on sentences you genuinely have to say lasts longer than hunting for published material.

The sounds you can produce and the sounds you can hear sit close together

★Which of the two changes first is not something that can be shown here; there is no measurement to that depth on hand. Still, people who practise pronunciation often feel their listening shift before their speaking does. A distinction you cannot produce is a distinction you struggle to hear. ★Which of the two changes first is not something that can be shown here; there is no measurement at that level available. Still, people who practise pronunciation often feel their listening shift before their speaking does. A distinction you cannot produce is a distinction you struggle to hear. Once you can produce the two distinctly, they start catching your ear when somebody else says them., they start catching your ear when somebody else says them.

This is another reason to use a scoring tool. Raising the number is not the point. Finding out that you had been treating two sounds as one is. From the moment you know, the resolution of your listening starts to change.

So if a few days of practice have not changed how you sound, there is no reason to stop. Your ears move first and your mouth follows. If you want a read on progress, whether the other person's words catch your attention more than they used to shows up sooner than a score does.

How much is enough per day?

Coming back to it briefly, many times, with gaps in between, stays with you better than setting aside one long block. 5 minutes is fine. Short, daily sessions work better for this than one long session a week.

As a routine, 3 steps are easy to hold on to: hear a model so you know the correct sound; record the same sentence and have it scored; go back to the model for only the sounds it flagged. Chasing the number from the second step does less than increasing how often you do the third.

And once you can score well on words and short sentences, check the sound of yourself in a real speaking situation. Reading a prepared sentence carefully and speaking while figuring out what to say do not involve the same pronunciation skills.

Summary — which one to open, by purpose

• See your pronunciation inside real lessons → SpeakCoach (measures the sound inside conversation, and the recording stays)
• One measurement right now → AnyToSpeech (no account, no install; scores on 4 dimensions, and states that recordings are not stored)
• Something to keep doing every day → ELSA Speak (a learning path means you do not have to decide what to work on)
• Be asked to repeat yourself less often at work → BoldVoice (the official site names 3 things it scores: pronunciation, stress, clarity. Paid after the 7-day trial)
• See whether your phrasing is natural too → Speak (paid after the 7-day trial)
• Check the correct sound of a single word → Cambridge Dictionary (British and American, plus the transcription)
• Hear how it is pronounced in context → YouGlish (finds and lines up real video clips)
• Turn your own material into audio → NaturalReader (Free forever plan)
• A rough check on getting through, with what you already have → Google Translate (there is no scoring)

Worth remembering

• 4 kinds of tool come up under pronunciation checker: scores you, plays a model, makes a model, recognises only
• Of the 9 here, 5 score your own voice
• 4 things to settle before choosing: free with or without an expiry, the unit measured (word, sentence, conversation), how fine the scoring is, whether the recording is kept
• The free shelf has 7-day trials mixed into it. Look for an expiry date first
• Prices move with campaigns. Take the figure off the official site, not off a roundup
• More than the number, what works is finding out that a sound you were sure of is off
• Which syllable you press on and the rhythm of the sentence decide whether you get through, more than individual sounds do
• Scores rise when you repeat the same sentence. The score cannot tell you whether that was pronunciation or familiarity
• Come back briefly and often, with gaps. Once words score well, check how you sound in a real speaking situation