Choose input
Use speech, screen subtitles, or both together.
REAL-TIME TRANSLATOR
RTT
Watch films, streams, meetings, and voice chats with movable bilingual captions for what you hear or see.
Android 10+ · Windows 10 2004+ x64 · Local mode can work offline
Start by choosing the input you need
RTT does not play or alter video. It places an independent caption layer above your player or chat app.
Choose a route
Select local models, or choose separate cloud profiles for speech and text translation.
Start translating
Grant recording, overlay, and screen-capture permissions. Screen subtitles need a region selection first.
Built for three everyday situations
Speech and screen features are independent, so you only configure what a situation needs.
Video and live streams
Capture permitted app audio and show source text with Simplified Chinese captions.
Voice chats and meetings
Use the microphone to hear remote speech played through a speaker. Protected call audio cannot be captured directly.
Video with source subtitles
Select the subtitle area on screen. On-device OCR waits for stable text before translating, and can run with speech.
Start with a working model, then choose more
The app keeps model downloads and manual selection available. The package includes core offline capability; other models download only when you choose them.
Vosk small models
Speed-first. Download English, Japanese, and more single-language models as needed.
SenseVoiceSmall INT8
The default accurate local multilingual route for English, Chinese, Japanese, Korean, and Cantonese.
Omnilingual CTC 300M INT8
For broad language coverage. The new startup path fixes its factory issue without a model re-download; RTT uses language rules when no punctuation model is available.
Qwen3-ASR 0.6B INT8
Optional high-accuracy ASR, about 879 MB, for high-memory Android devices and Windows.
Local and dedicated translation
Use ML Kit packs, validated Marian ONNX packages, DeepL, Baidu Translate, or LibreTranslate for text translation.
Choose your platform
The release page contains APKs, Windows installer or portable packages, and matching SHA-256 checksums.
Universal Android APK
App audio or microphone input, floating captions, selected screen subtitles, model management, and a session transcript. Translation runs in a foreground service, so it continues after you leave the app; restart after reboot or capture authorization expires.
Windows x64
WASAPI output loopback, an always-on-top caption window, tray controls, and local models. The first release focuses on the default output device.
Your content stays on your device by default
RTT collects no telemetry. Diagnostics are off by default and never record audio, caption text, or API keys. Local mode uploads nothing; cloud mode identifies the active capability, model, and provider.
Know before you start
- DRM media, apps that prohibit AudioPlaybackCapture, and FLAG_SECURE screens cannot be bypassed.
- Protected Android call audio cannot be captured directly. For voice chat, use a microphone with speakerphone.
- Chinese voice output is delayed translated speech, not a lip-synced film dubbing track.
- Larger models are usually more accurate but need more storage, memory, and power.