Your microphone
LiveYou speakThey hear the result
Open-source desktop app · Beta
Speak in your language. Hear the conversation in another.
Real Translate listens to your microphone and system audio independently, transcribes each side, translates it and can speak the result through the output you choose.
Windows, macOS and LinuxSession-only transcript
One conversation · two independent directions
Your voice goes out. Their voice comes back.
Run either channel on its own or use both together. Translation, target language and spoken playback stay independent for each side.
System audio
LiveThey speak in a call or videoYou hear the result
Start in minutes
Set the routes once. Then just talk.
01
Connect securely
Sign in with your Orcestr account and save your OpenAI API key in the operating-system credential store.
02
Choose each direction
Select microphone or system audio, Transcribe or Translate, a target language and optional spoken output.
03
Start the session
Follow source and translated text live, copy what matters, or clear the session whenever you want.
Designed for a real call
Focused controls, native audio routing.
The app keeps the two sides understandable without turning a conversation into an audio engineering panel.
Translate or only transcribe
Translation is optional. Each channel can preserve the original speech as a live transcript instead.
Speak only when useful
Mute generated speech per channel while keeping recognition and translated text running.
Separate devices
Choose microphone input, system loopback source and a translated-audio output for each direction.
Speech-first input processing
Noise suppression, an adaptive noise floor, gain, compression and limiting help distant speech remain usable without amplifying silence into a false signal.
Temporary transcript
Text stays in memory while the app is open. Copy or clear it at any time; closing the app discards it.
Native on every desktop
WASAPI on Windows, ScreenCaptureKit on macOS and PulseAudio or PipeWire monitor capture on Linux.
Private by design
Your key stays in native secure storage.
Raw audio is handled by the Rust process and sent directly to OpenAI for the selected Realtime mode. The renderer receives bounded status, level and transcript events—not your long-lived API key.
OpenAI usage is billed by OpenAI to the account behind your API key. Orcestr does not proxy or resell that usage.
Platform behavior
System audio is handled natively.
Windows
Process loopback captures the selected system source while excluding Real Translate playback, so translated speech can return to the same physical output.
macOS 13+
ScreenCaptureKit captures system audio and excludes the app itself. The first session requires Screen & System Audio Recording permission.
Linux
Uses the default PulseAudio or PipeWire monitor. Because one client cannot always be excluded, choose a separate translated output or mute spoken playback for that channel.
Desktop beta
Download Real Translate
Installers are built from a tagged commit on the main branch, signed for the Tauri updater and published directly to immutable S3 storage.
Latest versionv0.1.5
Questions