Description of App
Listen2 runs neural text-to-speech directly on your iPhone or iPad without connecting to cloud services. No subscription or accounts — just a 7 day free trial and then a one-time purchase.
Most text-to-speech apps either use the tried and true system voices or require monthly payments for cloud processing. Listen2 runs neural voice models locally using your iOS device's CPU and Apple Neural Engine, giving you natural-sounding speech without ongoing costs or privacy tradeoffs.
WHAT MAKES LISTEN2 DIFFERENT
Multi-Voice Engines, One App
Piper provides reliable neural voices across 13 languages. Supertonic offers a fresh take on a voice engine — flow-matching models running on Apple's Neural Engine via CoreML for natural rhythm and expressive speech. Choose from 100+ voices in English, Spanish, French, German, Italian, Portuguese, Russian, Turkish, Greek, Hindi, Vietnamese, Swedish and Hungarian. More coming.
Voice Tuning
Adjust the model parameters of any voice to suit your preferences. Save presets to switch between profiles. Preview pronunciation rules before saving them.
Import Anything
PDF, EPUB, DOCX, DAISY, markdown, plain text, clipboard, or web articles. Share from Safari and Listen2 extracts the article text. Paywalled content works (even sites on which other major TTS apps fail) when shared from a browser where you are logged in.
Machine Learning Word Sync
CTC forced alignment matches audio to text at the word level. Watch each word highlight as it is spoken — across Latin and non-Latin scripts alike. Highlighting engine newly re-build to be lean and efficient that's dramatically easier on your battery life.
Smart Document Reading
Multi-column layout handling, footnote filtering (for PDF files), hyphenated word rejoining. AI Vision for scanned PDFs. Full EPUB, DOCX, and DAISY support.
Figure Descriptions
On-device AI vision generates spoken descriptions of figures, charts, and photos as you listen. For non-English documents, descriptions translate automatically via Apple's on-device Translation framework. (Works with EPUB and PDF as a first roll-out)
Pronunciation Rules
Fix how any word is spoken without changing display text. Rules for names, terms, or abbreviations. Live preview before saving.
Built-in presets for footnote markers. Global or per-document. Doesn't break highlighting.
True Offline Privacy
Processing, normalization, and synthesis all happen on-device. Documents never leave your phone. No uploads, account, or tracking.
FEATURES
- 100+ neural voices across 13 languages, plus dozens of Apple system voices
- 3+ engines: Piper (broad language support), Supertonic (high-fidelity CoreML, flow-matching models), and Kokoro. Apple system voices offer efficient, quality options if using enhanced or premium models.
- Voice tuning with saveable presets
- Word-by-word or sentence highlighting
- PDF, EPUB, DOCX, DAISY, TXT, Markdown support
- AI figure descriptions with auto-translation (PDF and EPUB support)
- Collections with intros, cover art, and auto-advance
- Web Article Collections — batch articles into an auto-playing queue
- Clipboard import and Siri shortcut support
- 0.5x to 2.5x playback speed (Kokoro supports up to 1.4x because of real-time constraint)
- Background playback with lock screen controls
- Import from Files, links, Share Extension, or other apps
- Export any document as a plain-text file from the library
- Pronunciation rules with live preview
- Atkinson Hyperlegible and OpenDyslexic fonts
- Full VoiceOver support with rotor navigation and Magic Tap
- Elapsed and remaining time in now playing & reader view.
WHO IT'S FOR
- Commuters turning drive time into learning time
- Students reviewing papers and textbooks
- Professionals handling confidential documents privately
- VoiceOver users looking for a fully accessible TTS reader
- Readers with dyslexia who benefit from audio-visual sync
- Anyone tired of TTS subscriptions
One-time purchase. 7-day free trial. No subscriptions.
Comments
TestFlight beta link
Here is the link to the TestFlight beta of this app I mentioned in the above comments.
https://testflight.apple.com/join/uRrWXCba
HarmonicaPlayers response
do you get charged for the voices if you use the testflight beta or the voices all free
TestFlight beta
Hello,
Because of the way TestFlight betas Work you should not be charged when it asks you to purchase. You can make sure of this by swiping through the screen for payment. It should say something like you won’t be charged. For the App Store version you will get a seven day free trial then if you decide you want to keep the app pay $24.99 once and you’re done forever. No charges for voices no charges for subscriptions just a one time purchase. Hope this helps clarify things.
Bookshare
What is the easiest way to get books from bookshare to read in this application?
Adding Book Share books to Listen2
Because I use multiple reading applications I often download books from the Book Share website directly then save them in my iCloud Drive when I import documents from my Files application they should pop up immediately in your recent. The developer has reached out to Book Share to ask about becoming a technology partner, but as far as I know has not received a response. In the meantime downloading books to your Files application and importing them as you were a traditional E pub or PDF file is the fastest way. this does support Daisy books it will not matter if you download them with or without images they will still import.
Some Feedback
I like the ap and think it has potential. However, listening to a document on my iPhone 15 Pro, I notice that this app makes the phone extremely hot. I assume this is from processing all of the AI for the voices. Also, when browsing voices, how about grouping them together and giving them coherent voice names? Most of the voice names make no sense. Overall, a nice beginning though.
Phone heating
I posted this in one of the forums recently and will post it here as well for the sake of argument. Unless you require highlighting while listening, switch it to sentence mode, or turn it off. This will decrease the overheating potential. My default the app uses word level highlighting. As to the voice names, those voice names are what are tied to the Piper models voices. If you use the add-on for NVDA, you will see the exact same voice names in their voice manager.
Questions and comments and answers - from Zach
HarmonicaPlayer - The app is a one time purchase of $24.99 after a 7 day trial. The individual voices are all included with that single purchase. There are no other in-app purchase — all voices are included. But, blind-harper is correct. If you have the app installed with TestFlight, you can go through with "purchasing" at the pay wall and it will NOT charge you. You have 90 days to test per build in TestFlight.
Chad - I'll follow up with Bookshare. Someone at Bookshare forwarded my request to the right person, but haven't heard back yet. But yes, blind-harper is correct that getting those book files into files is a good way to import them. If the files are pdf, epub, daisy (including zip files), or docx, you should be able to "share" those files with Listen2. But the book files are already saved to files, you can start from Listen2 and tap the plus icon in the upper right and choose "import file" and then navigate to the file from there.
sockhopsinger - Yes, the heating up is definitely a thing especially with text highlighting set to word. You have 2 machine learning models processing each text chunk real-time. Now with v1.5 (just released 2 days ago), there was a pretty good efficiency improvement and may be worth trying again. But like blind-harper mentioned, if you are not using word highlighting, set text-highlighting to either sentence or off and that will save half the processing your phone is doing. Also note that as we speak, I'm training very tiny, efficient word alignment models per language and they run on apple neural engine. So if everything goes well, I'll have current word alignment model out of the app and include these tiny efficient ones on the next release. This should make the heat problem go away altogether from my initial testing.
sockhopsinger - Also, about the voice library - I just added some extra filters. Up in the top right corner, there's a new filter button that opens a sheet that lets you choose how the voices are sorted, which voice engines are visible (piper, supertonic, system), which tier of system voices are visible, which languages are visible, as well as a search bar for searching for voices by name. I'm hoping that makes finding the voice you're looking for better. But yes, as blind-harper already mentioned, the Piper voices are just using the name straight from the piper project. I do take that as feedback about those voice names and might consider something like letting the user add a custom name for a voice.
blind-harper - thank you for fielding these questions and comments! I hadn't been getting notifications on this thread. I believe I will now that I'm leaving a comment.
-Zach
Listen2 Developer
Phone heating addressed
sockhopsinger - version 1.6.0 of Listen2 was released today and it has much better efficiency with word highlighting on. I trained super lightweight word alignment models per voice engine and language combo. I tested it on my iPhone 15 Pro Max by playing a book for over an hour and I didn't notice any excessive heat. Just wanted to mention this improvement in case anyone was concerned by the heat issue - and that it's much improved if not eliminated now.
-Zach
A Few Feature Questions
Hi. So by the time I found this app, which is awesome, by the way, I guess this "Supertone" discontinued their voice cloning feature? I'm curious if there will ever be another way to clone voices or to add them from other sites. It's not something that's a high priority to me though, and I'd rather just not have to pay yet another subscription.
One thing I am more curious about though, is there a good way to get a "multiple language" voice to switch back and forth in the same document?
And is there a way to export the audio rather than just the text?
Multi language voices
Hello,
From what I’ve been able to observe with the way the Supertonic voices behave is that as long as the language itself is clearly marked. Let’s say you have a document mostly in English but it also contains Japanese text as long as the app registers the Japanese text clearly it switches the language automatically. If you’re reading documents in another language you select your base primary language and the voice will switch from whatever your primary language is to other languages and mostly the same way. Note that this is only predicated on observable behaviour between Spanish, English, and Japanese but if it does it with those three I would like to think it does it with the others this model supports. As to voice cloning there are some get hub repositories that are looking to re-create the voice builder experience one of them is a Google Kola notebook but to do it appropriately you either have to know how to ask the coding side of an AI platform to create a Supertonic compatible voice clone based on an uploaded recording, which is what I did through the work section of ChatGPT where it pulls in all the relevant dependencies and training data sets or using one of the ghetto repositories doing it manually yourself. What helps in my case is that I had actually purchased a voice builder clone of my voice and could use it for part of my coding project and essentially drive another clone from that voice clone when I wanted to use a slightly different set of recordings.
Thank You
Thank you because when it doesn't work it really makes me cringe. lol