Guide

How to add live subtitles to any app on iPhone

iOS does not let one app read another app's audio. The only route that works is to start a screen-sharing session and float the subtitles in a Picture-in-Picture window on top of whatever you are watching. That is what Hono Subtitles does: it listens to the audio your other apps are playing, translates it as it is spoken, and floats the result over the screen. It needs iOS 26 or later, and both the recognition and the translation run on the device.

Why this is hard on an iPhone

On a Mac you would solve this in ten minutes with a virtual audio device. Nothing like that exists on iOS, and that is deliberate: an app cannot see or hear what another app is doing. The sandbox is the feature.

So there is no "give me the audio of whatever is playing" API. It is not deprecated and it is not private — it was never there.

What does exist is the broadcast mechanism, the one built so you can stream your screen to Twitch or record a demo. It is the only sanctioned way for one app to receive a system-wide audio stream. That is the whole reason a system sheet appears when you start, and the reason the button on it says Start Sharing rather than anything about subtitles.

The second half of the problem is showing the text. Putting anything on top of another app means using the one primitive iOS has for it: Picture-in-Picture, which was designed for video playback. There is no overlay window on iOS, no floating bubble, no always-on-top layer.

Three things follow from that, and they are worth knowing before you install anything:

• You cannot screen-record at the same time. iOS has a single broadcast session. A subtitle app that works this way is using it.
• The floating window cannot be tapped. Picture-in-Picture is an output surface — it displays, it does not take input. Every control lives back in the app.
• Opening the camera closes it. The camera takes over the same surface, so the window is torn down. Picture-in-Picture survives over apps you watch things in, not over apps that use the camera.

There is one more constraint that does not show up in the interface but shapes everything behind it: the process that receives the broadcast has a 50 MB memory ceiling. Speech recognition and translation both have to fit inside that, on a live stream, without falling behind.

What "share my screen" actually means here

The sheet says screen sharing because that is the mechanism's name. It is not a description of what this app does with it.

What gets used is the audio stream. The broadcast hands over video frames as well, and the code that receives them does nothing — the branch is empty. The same is true of the microphone branch: the app never asks for microphone permission at all, so in-person conversation around you is not captioned, and not captured either. The only permission it requests is speech recognition.

Nothing is recorded, stored, exported or uploaded. Recognition runs on Apple's on-device speech models and translation on Apple's on-device translator. The first time you use a language pair, its models download; after that it works with no network connection at all. More detail is on the privacy section of the main page.

How to set it up

1. Start whatever you want to watch and leave it playing.
2. Open Hono Subtitles. Choose the language being spoken and the language you read.
3. Tap Start Sharing Audio. The iOS sheet appears — confirm with Start Sharing.
4. Tap Open Subtitle Window. The floating window appears.
5. Swipe back to what you were watching. The window stays on top, and lines appear as people speak.

The order of those last two is not a style choice. The sharing session has to exist before there is any audio to caption, so the window has nothing to show until it does.

What it works over, and what it does not

It works over anything on the device that plays sound: streaming apps, live streams, podcasts, video in a browser, recorded lectures, a meeting you are listening to. The app does not need to know what it is — it is reading the audio the system is playing.

What it will not do:

• Video calls where your camera is on. The camera closes the floating window.
• Sound coming from somewhere else. A laptop, a TV, a person in the room. It captions what this iPhone is playing, not what it can hear — there is no microphone involved.
• Screen recording in parallel. One broadcast session, already in use.
• Every language in both directions. The list is read from iOS at runtime rather than hardcoded, and the two sides are not the same list. Listening is limited to the overlap between Apple's on-device speech models and its translation models; reading is anything Apple's translator handles. A language can be available to read but not to listen to — Thai, for instance.
• Anything before iOS 26. The on-device recognition and translation models this depends on arrived in that release.

Questions