Voice & transcription
Speech recognition has stopped being a differentiator — Gemini, ChatGPT, and Google Recorder give it away free — so value has shifted from "transcribe" to "reliably catch the moment and hand the result back without friction." The winner isn't the one with the most accurate model; it's the one people trust with a court recording, a job interview, and a lawyer meeting, and the one that doesn't hold their data hostage.
Three findings
Capture you can trust
The recorder's basic job fails silently and catastrophically: empty files, silence instead of an interview, cutoff at the three-minute mark, an archive gone overnight. It hits the most vulnerable use cases — court, work, illness, evidence. Here, reliability isn't a feature; it is the product itself, and this is precisely where leaders lose trust permanently. Local backup, a live indicator confirming recording is actually happening, and an archive safety guarantee are an obvious unoccupied moat.
Paywall after recording is the most infuriating pattern in the niche30
The user records an entire meeting or lecture, waits through the upload, and then sees only 3–5 minutes of transcript behind a paywall. Their content is held hostage at the exact moment when redoing it is impossible.
I just recorded a full meeting, only to find out afterwards that I have to pay to get the full transcript after ish 15 min.
It did not tell me that it was only free for up to 5 minutes until AFTER I had already been talking and transcribing for a long long time.
Imagine getting 120min free of transcription only to find out you can only see 3 minutes of the actual transcription
Silent recording eats the audio — silence instead of an interview22
The app's core job — reliably capturing audio — fails without a sound: the user believes they're recording, then discovers an empty file or 45 minutes of silence, with no chance to redo it.
It recorded nothing. I have 45 minutes of nothing now so thank you for that.
Then I did an interview, and it recorded the whole thing in silence. They couldn't retrieve it, so I lost a 31 minute interview
It failed me when I needed it the most. I had to submit recorded calls to my employer, and now I don't have a recording, and pretty much I'm out of my job.
Long recordings cut off at 3–5 minutes without warning16
The recorder quietly stops after a few minutes, clips the beginning, or finalizes the recording the moment you switch away from the app. For lectures and long meetings this makes the product useless exactly where it matters most.
it doesn't keep recording in the background for very long, so I never know if my lecture or conversation or voice note is going to be incomplete
Has problem to indeed write long speech. 5' talk and captured fragments...
I noticed it after, that it stopped recording after 3 minutes.
"Why does a voice recorder need my work email and full calendar?" — trust collapses at onboarding16
To record a simple note the app demands sign-in with a work email, calendar access, and contacts — then proceeds to gate-crash others' calls and spam contacts. For a voice recorder, this reads as surveillance.
Why does a voice to text app need to see my life online? Nope, deleted.
otter will email everyone of your contacts or u have meetings with l telling them YOU invite them to otter, with reply to..YOUR email.
why do i need to create an account and give information for an app to record calls, now uninstalled before i even used the recorder
The recording archive vanishes overnight for no reason14
People build up years of songs, ideas, and legal evidence — then after an update, or simply one morning, everything is gone. This destroys trust in the product as a memory store more thoroughly than any other bug.
I've recorded 18 songs that I have written. then one day I logged on and everything was just gone. don't trust this app.
You have cost me not only amazing memories, but also possibly a court case.
I opened the app today to record a new one I had came up with, recorded it and went to listen to a few of the older ones I had. They're just all gone, every single one of them.
Trust as a product: court, job, illness, divorce — the cost of failure is enormous13
When it works, the product changes lives: it remembers for a long-COVID patient, keeps divorce records, relieves anxiety. The emotional attachment is the strongest in the niche. Those same people use the recorder in critical moments — and failure is not an option. Losing an interview, silence where a lawyer meeting should be, a vanished archive hit the users who need the product most. Reliability here isn't a feature — it is the product.
As someone suffering chronic illnesses, I struggle remembering my long, complex medical appointments.
I have long COVID & have short term memory loss & ADHD. I use it to record every meeting
Primarily for meetings with my attorney as I begin a divorce. The AI generation of notes and transcript have been amazing.
Accuracy and speakers as the real job
Transcription isn't needed for its own sake — people need an accurate meeting or interview record without manual cleanup. Engines drop words, merge speakers, output the wrong language, and can't handle bilingual meetings — exactly the cases users bought the product for. Whoever closes diarization, mixed-language support, and continuous-dictation punctuation will take the students, researchers, and international teams abandoned by the current leaders.
Accuracy falls short: dropped words, confused speakers, manual cleanup required28
The core job — an accurate transcript — is done poorly: whole phrases go missing, errors stack up every sentence, and the user spends more time correcting than they saved. That was the entire reason for getting the app.
the speech recognition is so bad you will be lucky to half your words right. it would be better just to type your dictation then to use this app.
it had a lot and I mean a lot of mistakes like 3 every sentence
Transcription not accurate. Only let's you record 30 minutes of lecture. Only had 1 speaker, but labeled it with multiple speakers.
One language per file — bilingual meetings fall apart18
Real meetings and lectures often mix two languages, yet the engine handles only one at a time: English terms inside Chinese speech turn into garbage, and accents or non-English speakers are recognized poorly. A huge segment left unserved.
When you choose Chinese transcription, all of the English words just become transcription errors and vice versa. It's a massive limitation.
My lecturer most of the time has a mix of 2 languages (bilingually) in his teaching. it could really help if the app is able to improve and have that feature.
I can't transcribe if we have multiple languages in on conversation. like minute of Enlisha amd Chinese.
Speaker separation breaks — the real unsolved job for meeting transcription14
Diarization is the key to transcribing meetings and interviews, yet apps merge two people into one, invent five speakers from a single voice, and give no way to correct the count. Without this, a meeting transcript is useless.
It runs the end of someone talking into the beginning of someone else talking into the same person. it also confuses people's voices, sometimes even men and women with dramatically different tones
Wrong people gets assigned to some lines in every transcription. There's no way too add more people when it gets the number of people wrong
seems limited. most of the time it can't distinguish between 2 speakers talking so it combines them together, getting it wrong.
Dictation breaks the flow: pause every 5 seconds, no punctuation, profanity censored12
For continuous thought-dictation the engine stops every few seconds, adds no punctuation, and even bleeps profanity in private notes. This breaks the entire "I talk, it writes" scenario the product promises.
it will pause after every 5 - 6 seconds and break your flow completely, don't download if you want to use it without premium.
If I or anyone speaking in the vicinity of my phone says a "bad" word, that word will be bl***ped. Or in some cases, not transcribed at all.
Reported back in 2021 that the app censors slurs, nothing ever changed. I'm using this as an author, and expletives are needed.
Transcript comes out in the wrong language — gibberish instead of text9
An English meeting comes back in Welsh, Afrikaans gets transcribed as German, the language setting is silently ignored. The user receives coherent text in a foreign language and realizes it far too late.
The first recording was entirely in Welsh (we're English). When I tried to translate it, it was gibberish.
broken. My meeting was in English and the transcript came out in some languages I don't know.
Started off great, now transcribing in Welsh even with the setting set to English
Honest economics and data freedom
The strongest negative sentiment doesn't come from quality — it comes from predatory mechanics: paywall after recording, card required upfront, charges after cancellation, transcripts locked inside the app, subscription on top of purchased hardware, forced five-star ratings. Against a backdrop of free AI alternatives, all of this pushes users straight to competitors. Transparent pricing, an honest trial on a real slice of your own recording, and frictionless bulk export aren't marketing — they're the primary driver of retention and word of mouth.
"Free trial" = card upfront and charges after cancellation20
They ask for card details before you've used the app, charge the annual fee immediately after the "trial," keep billing after cancellation, and the subscription doesn't appear in Google Play so you can't stop it. A dedicated churn and chargeback pipeline.
Why would a free trial NEED my credit card information?? It doesn't make sense. Its NOT FREE if they need a form of payment method.
I subscribed believing there was a free trial, but the annual fee was charged immediately. I canceled the same day and did not use the service.
I canceled the trial but was still charged. I tried multiple times to contact Customer Service but all they have is a chat bot that does not work. I will be disputing the charge.
Transcripts can't leave — text is locked inside the app17
The "copy summary" button was removed, bulk export doesn't exist, you can only download one recording at a time out of hundreds, and sharing means a link only. Users feel trapped in someone else's software holding their own data.
they refuse to allow mass export of your recordings or transcripts so you are locked into their terrible software. I finally wrote a script to get all my stuff out
cannot copy meeting summary out of app. before you uses to be able to copy the summary of the meeting. now you can only export or copy the whole transcript.
to not have a copy/cut function on an app is horrible. needs the ability now!
Call recording requires dialing a service line and merging — it exposes you and it's maddening15
When it works, it captures both sides cleanly — something free apps can't do. To start recording you have to call a service line and merge the calls — the other person sees "conference call," hears tones, and has already hung up or grown suspicious by the time recording begins. The mechanic gives away that you're taping them.
The caller you are trying to record sees the message conference call and it makes me sound far away to them.
there is an approximate 30 second delay before the merge function is available to actually begin the recording for each call. By the time I can get a recording started they've either hung up or become too suspicious
because you have to call another line and merge them in order to activate the recording it can make it easy to detect that you are up to something.
Hardware purchase plus subscription on top plus a minute cap = outrage13
Those who are used to it praise it as a "pocket assistant": transcript and summary ready by the time they're back at their desk. After buying the expensive device, users discover that AI features and any reasonable minute allowance require a separate subscription, and 300 free minutes amounts to fewer than 5 lectures. It feels like paying twice for something a phone does on its own.
Without the subscription, you can only use YOUR device, the device YOU paid for, 300 minutes, or 5 hours per month. That is less than 5 classes. Returned.
you can record and transcribe with software that comes free with your phone and then use your own ai.
Forcing users to buy a device is the WORST business strategy.
Free Gemini, ChatGPT, and Google Recorder are named as direct replacements12
Users openly write that they upload audio to Gemini or ChatGPT and capture with the free Google Recorder — getting the same or better results for nothing. A paid transcription engine stops being a value proposition on its own.
upload your sound file into gemini it transcribes it flawlessly in seconds. Then just copy it onto your clipboard and paste it any where you want.
exorbitant prices which you can now just use free AI servers to achieve the same thing.
I discovered chatGPT does a very good job when i play spoken Cantonese video, & change it to text. ChatGPT can got the words right with ~95%
Forced 5-star review and pornographic ads at launch — instant uninstall11
The app locks features until you give five stars, and runs porn-adjacent ads on the start screen. People delete it without trying, warn others about children, and it sends a powerful negative signal across the entire budget-transcriber category.
They offer you a free plan then tell you to rate them 5 stars to access it. Im a woman of God. I cant rate before I use. I cant lie.
di paksa untuk kasih bintang 5 sblm bisa pakai apps ujung2nya di paksa bayar! scam!!!
Deleted based on ads, never even tried it out. Their ads are porn ads. Be careful if you have kids.
Two more findings — with the breakdown and review quotes.
7 opportunities
Ideas users ask for themselves — each backed by proven demand.
10 apps
The leaders — Otter, Notta, Plaud, Transkriptor — are built around retention: trial traps, 3–5 minute limits revealed only after recording, locked export, subscriptions stacked on top of hardware. That generates revenue but creates a trail of hatred and explicit callouts directing users to free AI alternatives. Reliable recorders (Easy Voice Recorder) and honest dictation apps (Speechnotes) lose users to vanishing archives, long-recording cutoffs, and mediocre accuracy — core-job failures, not AI feature gaps. Nobody has closed the trifecta of reliable capture, honest monetization, and frictionless data export.