Could Thoth come to iPhone? What I found
2026-05-18
The most common feature request I get is "iPhone version," closely followed by French localization. I have been putting the iOS one off because I assumed it would be too limited for what Thoth does. So I finally sat down and mapped it out properly.
The short answer is: more is possible than I thought, but one fundamental limitation changes what the app would be.
The one thing iOS cannot do
Thoth on Mac records system audio. That is how it captures Zoom calls, Teams meetings, Google Meet, any audio playing on your Mac. This is a macOS-only capability. There is no equivalent on iPhone. No entitlement, no workaround.
This is not a limitation I can engineer around. It is a platform boundary.
So an iPhone version of Thoth would record only your microphone. That makes it a completely different tool. It is useful for in-person meetings, interviews, lectures, voice memos. It is not useful for remote calls, which is the primary use case most people have.
Any iOS version would need to make this clear upfront, and point users toward the Mac app for remote meeting recording.
Background recording works fine
The good news is that microphone recording in the background works well on iOS. If you start a recording and lock your phone or switch apps, the recording keeps going. iOS handles this with a background audio mode that keeps the session alive.
There is one visible side effect: the orange dot in your status bar. iOS shows it whenever any app is recording your microphone in the background. You cannot hide it and you would not want to. It is a privacy feature.
The trickier issue is interruptions. Phone calls, Siri, FaceTime, and alarms can all pause your recording mid-session. The app would need to handle these gracefully and send a notification when it happens, so you know the recording was paused and can decide what to do. Something like: "Recording paused because of an incoming call. Tap to resume." That is doable with local notifications on iOS, no server involved.
Transcription: better than expected
I assumed iOS would force a step down to tiny Whisper models and nothing else. That is not the case.
WhisperKit, which powers transcription in Thoth on Mac, has full iOS support. On recent iPhones it can run models up to Whisper Large v3 Turbo, which is a roughly 1 billion parameter model. This is real transcription quality, not a watered-down mobile version.
The bigger surprise: Parakeet, the newer and faster transcription model I am integrating into Thoth on Mac, also runs on iOS. FluidAudio supports iOS 17 and above, and there are already real apps shipping with it on iPhone. The same multilingual batch transcription and live streaming that makes Parakeet interesting on Mac would be available on iPhone.
Older devices would fall back to smaller models automatically. But on an iPhone 15 Pro or anything in the iPhone 16 line, transcription quality would be close to the Mac experience.
Running AI on iPhone
This is the part that surprised me most.
Apple announced a developer framework at WWDC 2025 called Foundation Models. It gives apps direct access to the on-device Apple Intelligence model, roughly 3 billion parameters, running entirely on your device with no server call. You get summarization, extraction, and rewriting for free. It is available on iPhone 15 Pro and all iPhone 16 models.
For AI summaries and enhancements, this would be genuinely useful. The quality is not at the level of a frontier cloud model, but it is solid, it is private, and it costs nothing to run.
For devices that do not support Apple Intelligence, there is still the option of running a custom downloaded model using Apple's MLX framework or llama.cpp. On an iPhone 15 Pro or newer with enough memory, models in the 3 to 8 billion parameter range are viable with 4-bit quantization. Slower than a Mac, and you will feel the phone get warm, but it works.
Older devices or lower-end models might not support on-device AI at all. The app would need to detect this and tell you clearly what is and is not available on your specific phone.
My plan would be to offer both: Apple Intelligence for supported devices, and a custom model picker for users who want more control or support older hardware.
The idea I am most excited about: the iCloud bridge
Here is where it gets interesting.
Imagine recording a meeting on your iPhone, then opening Thoth on your Mac later and re-transcribing with Parakeet, running speaker diarization, and getting the full AI summary, all from audio that was already waiting there via iCloud. That is what an iCloud sync setup could look like.
The iPhone captures the audio and produces a live transcript using whatever model fits the device. The recording syncs to iCloud. When you open Thoth on your Mac, you can run re-transcription and AI enhancement on it there, and the result syncs back to your iPhone.
This solves the hardware problem neatly. If your iPhone is too old for on-device AI, your Mac handles it when you get home. The iPhone becomes a high-quality capture device, and the Mac is the processing engine when you need the full treatment.
The privacy story here matters. Your audio would move from your iPhone to your iCloud to your Mac, all within your own Apple account. Thoth's servers are not involved at any point, because Thoth has no servers. iCloud Drive transfers are encrypted in transit, and with Advanced Data Protection enabled on your account, the files are end-to-end encrypted at rest as well. This is a very different posture from meeting recorders that upload your audio to their own infrastructure to process it. The iCloud bridge would keep everything inside the Apple ecosystem you already trust, without introducing any new party into the chain.
It also means the app library would be shared. Recordings made on Mac show up on iPhone. Recordings made on iPhone show up on Mac. One app, two entry points.
Pricing across platforms
One thing I feel strongly about: if someone buys Thoth Pro, they should not have to buy it again for iPhone. Purchasing on one platform would unlock both, using StoreKit's cross-device entitlement system. Apple supports this natively, so there is no custom infrastructure needed.
Where this leaves things
I am not announcing an iOS app. This was an exploration, and there is a lot of work between here and a real product. But I learned enough to know it is worth pursuing at some point.
The things I thought would be hard, transcription and on-device AI, turned out to be more tractable than I expected. The fundamental limitation, no system audio capture, is real and would define what the iOS app is. It would be a companion to the Mac app, not a replacement.
If you record a lot of in-person meetings or interviews on the go, it could be exactly what you need. If your primary use case is recording remote calls, you will always need the Mac.
I will keep this on the roadmap and see how the Mac version develops first.
Thoth is a private meeting recorder for Mac. All audio stays on your device. Built by one person, no funding, no team. If you find it useful, upgrading to Pro is the best way to support development.