SyncTrans 1.0.0 has been out for a week, and the question we get most is the simplest one: what does using it actually look like? This post walks one complete session on a shipping build — the scenario is a Chinese-speaking client meeting held in English, two people sharing one MacBook on a table. Every step below is what 1.0.0 does today; the longer reference version lives in the user guide.
Setup: five screens, all skippable
Install is the usual DMG drag into Applications. The first launch opens a five-step setup: languages, plan, display and audio, model combination, and a ready screen that connects the service. Two things about it are worth knowing before you start clicking. First, every step has a working default and the whole setup can be skipped — none of it is a gate, and every choice reappears later in Settings. Second, the setup runs once; interrupt it at the last step and the next launch still takes you straight to the main window rather than making you walk it again.
For this session we picked the managed cloud on the plan step. That choice needs no account: the ready screen claims free starting credit for a guest session and starts the service, so the distance between opening the app and saying the first sentence is genuinely a few minutes. Registering an account afterwards adds its own free grant on the same Mac, which is the intended behaviour, not a loophole.
Languages: one main, one target, two speakers
The pair for this meeting is simple: Chinese main, English target. The setting that does the real work is in Settings → Sound — the second-language option on the microphone channel. With it set, completed turns from the one built-in microphone are attributed to the Chinese speaker or the English speaker by the language they were spoken in. There is no second microphone and no button to press before talking; the transcript keeps the two people apart on its own. The two languages must differ, which is the only way to misconfigure it, and the app says so rather than accepting a same-language pair.
The conversation itself
Starting is one button: Start translation, or ⌘R. The first capture triggers the macOS microphone prompt; allow it once and it never appears again. If you deny it by reflex — we did, on the first take — the failure is not silent: the status capsule in the toolbar flips to “needs attention” and a banner above the conversation carries an Open Sound Settings button that lands on the right privacy pane. Fixing it took under a minute and cost no session state.
While the session runs, each half-finished sentence shows as a dimmed draft line and is replaced by the committed turn a moment later. Drafts are never translated and never enter the history, which matters more than it sounds: it means the record only ever contains sentences the app actually stood behind. When someone trails off mid-thought and the endpoint logic waits, the complete current sentence button at the end of the draft line (⌘Return) commits the turn immediately through the same pipeline.
Read-aloud ran on the managed voice for English and a second voice for the Chinese renderings, both chosen by name from the picker with a preview button — no identifiers anywhere. With one pair of headphones shared across the table, split-channel output sent English to the left ear and Chinese to the right; the two ears are separate playback lanes, so overlapping translations speak at the same time instead of queueing.
Afterwards: the record
Everything was already saved — every turn is written to the current project as it happens, so stopping the session loses nothing. We fixed one misheard product name in the history before exporting; the edit changes the record and deliberately does not re-run the translation behind your back. Export produced a parallel Markdown file with speaker and time per turn, which went straight into the meeting notes. Plain text and timed SRT are the other two formats, and none of the three involves an upload — the history lives on the Mac whichever engine produced it.
What we would tell a first-time user
Three habits cover most of the difference between a smooth session and a confusing one. Check the status capsule before the room fills up — its four rows (microphone, recognition, translation, read-aloud) name the stage at fault instead of leaving you to guess. Confirm the live input in the microphone popover; the level meter next to each device shows the room mic reacting before anyone says anything important. And put the two or three names the meeting turns on into the glossary first, because those are the words a general model will reinvent on every turn.
SyncTrans 1.0.0 needs macOS 14 or later, and you can download it here. The full task-by-task version of this walkthrough is the guide, and the detailed feature reference is on the Features page.
Questions or bug reports? The feedback board is the fastest way to reach us.