Nothing leaves the conversation
No account, network client, cloud transcription, analytics, or telemetry. Your words remain on the device in your hands.
On-device by default
iOS · iPadOS · macOS 26+
Two people.
Two languages.
Easy2say listens through one microphone, keeps both people visible, and drafts bilingual captions on device — without accounts, cloud transcription, or telemetry.
Equal space for both languages
Source and translation each receive half the screen: stacked in portrait, side by side in landscape.
A screen made to share
Flip the other person’s half across a table, or let landscape mode open the conversation side by side.
01 · The experience
Designed around the person across from you.
Not a transcript squeezed into a utility window. A calm, shared surface for following each other in the moment.
Live bilingual canvas
Original speech and its translation stay visible together, with a clear half of the screen for each.
Two-way conversation
Two people can speak their own languages into one microphone and read the other person’s words in theirs.
Across-the-table view
Flip one half 180 degrees so both people can read naturally with the device placed between them.
A graceful noisy-room override
The app compares both language lanes; if noise confuses it, tap either half to say who has the floor.
Local Taigi captions
A bundled 4-bit Breeze-ASR-26 model offers assistive Taigi-to-Chinese-character captions without a model download.
Transcript at hand
Review the session inside the app, then return to the live view without leaving the private workflow.
02 · How it works
From silence to shared understanding in three gestures.
- 1 Choose both languages Set the language being spoken and the subtitle language each person wants to read.
- 2 Set the screen between you Use portrait, landscape, or flip one half for an across-the-table conversation.
- 3 Speak naturally Tap Start. Speech and translation arrive together, entirely inside Easy2say.
03 · Language coverage
The list follows what your device can actually do.
Easy2say asks Apple Speech and Translation for the languages available on the current device. When a pair has no Apple Translation model, on-device Apple Intelligence may provide an experimental draft.
Speech and conversation examples
- Cantonese
- Chinese (Simplified)
- Chinese (Traditional)
- English
- French
- German
- Italian
- Japanese
- Korean
- Portuguese
- Spanish
- Taigi · Chinese-character captions
Additional translation examples
- Arabic
- Dutch
- Hindi
- Indonesian
- Russian
- Thai
- Turkish
- Ukrainian
- Vietnamese
The picker shows the exact choices available now; Apple may download a language asset before first use. Read the full technical notes ↗
04 · Privacy, literally
No account. No network client. No tiny print.
- Microphone audio and captions are never sent to an Easy2say server — there is no Easy2say server.
- Speech recognition and Apple Translation run on device with local language assets.
- For unsupported translation pairs, Apple Intelligence may draft an experimental translation on this device.
- Voice activity detection and the optional bundled Taigi model run through Apple’s Core ML stack.
05 · Build the app
From source to your iPhone in one quiet sitting.
The iOS experience is built from this fork. Bring Xcode 26, an iOS 26 device or simulator, and your own signing team for a physical device.
1
Prepare
Install Xcode 26, XcodeGen, and the Hugging Face CLI. A physical device also needs a local development team in Signing & Capabilities.
2
Fetch and generate
Clone the fork, fetch the pinned Taigi model, and generate the iOS project.
git clone https://github.com/audreyt/easy2say.git
cd easy2say
brew install xcodegen huggingface-cli
./scripts/fetch_breeze_asr_26.sh
(cd ios && xcodegen generate)
open ios/v2s-ios.xcodeproj
3
Run
Choose the v2s-ios scheme and an iPhone, iPad, or simulator, then run from Xcode.
Easy2say
Let both sides of the conversation stay visible.
Open source, on device, and ready to install on Mac.