Pocket TTS is now available on iPhone and iPad through ToBe SAID, bringing another modern offline AI system voice engine to iOS.
Until recently, Pocket TTS was available through ToBe SAID on Android and then on macOS, but not on iPhone. That has now changed.
The new iOS version of ToBe SAID brings Pocket TTS to iPhone, although Apple’s tighter memory and resource limits mean that it cannot use exactly the same higher-quality model available on the Mac.
Even with those compromises, Pocket TTS on iPhone is an interesting new alternative to Apple’s built-in voices and to Piper, which until now has been one of the main options for modern offline neural system voices on iOS.
Pocket TTS on iPhone Requires a Smaller Model
The macOS version of ToBe SAID demonstrated what Pocket TTS can do when it has access to the considerably greater resources available on a Mac.
Bringing Pocket TTS to iPhone has required compromises.
The developer has explained that iOS imposes strict memory limits on third-party text-to-speech engines. To make Pocket TTS practical on iPhone, the model had to be compressed further, reducing voice quality compared with the Mac and Android versions.
The developer has also experimented with somewhat larger iOS models in an effort to find a better balance between voice quality, memory consumption and stability.
As a result, users should not expect Pocket TTS on iPhone to sound exactly like Pocket TTS on a Mac.
Even with those compromises, however, the voices remain impressive. In my testing, I still find them better than Apple’s built-in voices.
You can find ToBe SAID for iPhone, iPad and Mac on the App Store. A separate ToBe SAID Pro edition is also available.
Pocket TTS vs. Piper on iPhone
The arrival of Pocket TTS on iOS makes the comparison with Piper particularly interesting.
Piper already provides modern offline neural system voices on both iPhone and Mac, with essentially the same app and voice technology on both platforms.
ToBe SAID takes a different approach. On Mac, Pocket TTS can use a higher-quality implementation. On iPhone, it has to use a smaller and more constrained model designed to stay within Apple’s mobile resource limits.
This means that Pocket TTS on iPhone does not preserve the same voice-quality advantage over Piper that ToBe SAID can have on the Mac.
However, Pocket TTS appears to have another potential advantage: latency.
In my testing, ToBe SAID feels at least as responsive as Piper and possibly somewhat more responsive. I have not performed a controlled latency benchmark, so this should be treated as an observation rather than a precise measurement.
For a normal listening application, that might sound like a relatively small difference. For a system text-to-speech engine, especially one used with accessibility features, it can be much more important.
Why Pocket TTS Latency Matters for VoiceOver
When comparing AI voices on iPhone, it is tempting to focus almost entirely on how natural each voice sounds.
For a system TTS engine, voice quality is only part of the story.
Latency matters too.
During continuous reading, higher synthesis latency may occasionally appear as an additional pause between generated segments. How noticeable those pauses are depends on the hardware, the TTS engine and how the application buffers and queues speech.
In a reading app such as Speech Central, speech normally consists of relatively long stretches of continuous audio. Small differences in synthesis latency may therefore appear mainly as occasional pauses or interruptions.
With interactive accessibility software such as VoiceOver, the situation is very different.
Moving between controls, characters, words, menu items and other interface elements repeatedly generates short new utterances. Any startup delay in the voice is therefore exposed again and again.
A delay that may be barely noticeable while listening to an article can make interface navigation feel considerably less responsive.
For VoiceOver users, the potentially low latency of Pocket TTS on iPhone may therefore be one of its most important advantages.
Piper Has One Major Advantage: Word Highlighting
There is also an important limitation of Pocket TTS that points in Piper’s favor.
ToBe SAID and Pocket TTS do not currently provide the word timing information required for synchronized word highlighting.
This is not merely an iPhone limitation. The same limitation applies to ToBe SAID on other platforms.
Piper, by contrast, supports the timing information needed for synchronized word highlighting.
This can be important for people who visually follow text while listening. In Speech Central, synchronized highlighting helps users maintain their position in a document while speech is playing.
It can also be valuable for accessibility, language learning, focused reading and workflows where visual reading and listening are used together.
This creates an important trade-off between Piper and Pocket TTS on iPhone.
- Pocket TTS may offer lower latency, which can be especially valuable for VoiceOver and other interactive accessibility uses.
- Piper supports synchronized word highlighting, which can make it preferable for users who follow text visually while listening.
Neither advantage makes one engine universally better.
How Good Is Pocket TTS on iPhone?
The most obvious question is how much voice quality has been lost by shrinking Pocket TTS for iOS.
There is a noticeable difference compared with the Mac version of ToBe SAID, where the model can take advantage of much greater memory and processing resources.
On iPhone, Pocket TTS no longer has the same clear quality advantage over Piper.
However, I still find the resulting voices better than Apple’s built-in system voices.
That makes the iOS edition particularly interesting. It is not simply a reduced Mac version. It represents a different balance between model quality, memory use and responsiveness.
Pocket TTS and Piper Are Both Fully Offline on iPhone
One of the most important characteristics of both Pocket TTS and Piper is that speech is generated locally.
There is no cloud TTS service generating every sentence, no requirement to maintain your own server and no per-character cloud synthesis bill.
This also means that modern neural voices can be used in applications that support Apple’s system TTS infrastructure without sending the text being read to an external speech service.
For people looking specifically for an offline AI voice for iPhone, both ToBe SAID/Pocket TTS and Piper are therefore significant options.
Using Pocket TTS with Speech Central on iPhone
Because ToBe SAID exposes Pocket TTS through Apple’s system text-to-speech infrastructure, compatible applications can use the voices directly.
That includes Speech Central.
Speech Central can use compatible system voices while providing the rest of the reading environment, including web articles, PDFs, books, documents, playlists, navigation controls and accessibility features.
In this setup, ToBe SAID provides the Pocket TTS voice, while Speech Central provides the reading application.
You can download Speech Central for iPhone and iPad from the App Store.
There is also a dedicated Speech Central for Mac.
Pocket TTS on iPhone and Mac: One ToBe SAID Pro Purchase?
The iPhone release is also interesting for people who already use ToBe SAID on Mac.
The iOS and macOS editions of ToBe SAID Pro appear to be distributed as the same App Store product across Apple’s platforms. In that case, an existing purchase should normally be available on the other supported Apple platforms rather than requiring another purchase.
In my testing, however, I already owned ToBe SAID Pro on the Mac and the iPhone App Store still asked me to purchase it again.
This appears to be a purchase-recognition or launch-related issue, although I cannot determine whether the problem is on Apple’s side or in the current App Store configuration.
If you already own ToBe SAID Pro on Mac, it is therefore worth checking your purchase status before paying for the iPhone version again.
Pocket TTS or Piper: Which Is Better on iPhone?
There is no universal winner.
If synchronized word highlighting is important to the way you read, Piper currently has a significant advantage.
If you use a system voice for highly interactive tasks, particularly VoiceOver, Pocket TTS’s potentially lower latency may be more important than small differences in absolute voice quality.
For continuous reading, the comparison is more subjective.
Latency can still matter because it may affect pauses between synthesis segments, but those events are generally much less frequent than during interactive screen-reader navigation. Their severity can also vary substantially depending on the iPhone model and the application using the TTS engine.
Voice quality, word highlighting, responsiveness and personal preference can therefore lead different users to different choices.
Pocket TTS on iPhone Is an Important Step for Offline AI Speech
The bigger story is how quickly offline neural text-to-speech on Apple devices is improving.
Earlier, ToBe SAID demonstrated on Android that Pocket TTS could work as a practical system-wide neural voice. Piper then made modern offline neural system voices practical on iPhone and Mac. ToBe SAID subsequently brought a higher-quality Pocket TTS implementation to macOS.
Now Pocket TTS is available on iPhone as well.
Apple’s resource restrictions mean that the iOS version cannot reproduce everything ToBe SAID can do on a Mac. The smaller model sacrifices some voice quality, but it still produces impressive offline speech and appears to retain very good responsiveness.
For iPhone users, that means there is finally meaningful competition between modern offline AI system voices.
And choosing between them is not simply about which voice sample sounds the most natural.
Voice quality, latency, VoiceOver responsiveness, synchronized word highlighting, hardware performance and privacy all matter.
Related: Pocket TTS Comes to Mac: ToBe SAID Brings Offline AI System Voices to macOS
Downloads:
ToBe SAID – Pocket TTS system voices
ToBe SAID Pro
Speech Central for iPhone and iPad
Speech Central for Mac