Vosk 英语小模型
内置轻量英语流式识别；局部结果延迟低
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Verified
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- en
- Platforms
- android, windows, web
- Size
- 41 MB
- License
- Apache-2.0
Idle
Live captions
Run speech, screen subtitles, or both. The browser only captures a microphone or screen that you explicitly authorize.
Select Share screen or tab, authorize it, then select the subtitle region.
Full source text and translations appear here after recognition starts.
Translator
Translate text, documents, subtitles, and video with visible progress and downloadable results.
0
Translation result
Translate documents into downloadable results
Text PDFs can be extracted locally. Scanned PDFs need OCR. Office layout requires a native document translation API.
PDFText layer / OCR routeOfficeDOCX · PPTX · XLSXStructured textHTML · JSON · XML · XLIFFSubtitles and imagesSRT · VTT · ASS · PNG · JPG · WebP
Generate downloadable subtitles from video
Extract audio locally, then use the active speech and translation routes to create SRT, VTT, ASS, or TXT. Audio is uploaded only after you explicitly choose cloud ASR and confirm.
No file tasks yet.
API connections
Choose providers by workload. General text models, dedicated translation, speech recognition, OCR, and TTS keep separate profiles and are never treated as interchangeable.
Configured APIs0
Saved credentials0
Active translation API-
Active speech API-
Active cloud routesData is sent only when a remote profile is current and its capability is running.
Capability profiles
ASR, translation, punctuation, OCR, and TTS are selected independently. Keep multiple named profiles for each capability.
Model directory
Android, Windows, and Web share the same capability package catalog. Web only loads runtimes explicitly marked for browsers.
Browser model storageOPFS · 0 MB
Ready models0
Desktop runtimeDisconnected
Web · Verified
Vosk 日语小模型
内置轻量日语流式识别；局部结果延迟低
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Verified
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- ja
- Platforms
- android, windows, web
- Size
- 50 MB
- License
- Apache-2.0
Vosk 中文小模型
轻量普通话离线识别；适合低内存设备
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- zh
- Platforms
- android, windows
- Size
- 44 MB
- License
- Apache-2.0
Vosk 韩语小模型
轻量韩语离线识别；适合低内存设备
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- ko
- Platforms
- android, windows
- Size
- 87 MB
- License
- Apache-2.0
Vosk 西班牙语小模型
轻量西班牙语离线识别；适合低内存设备
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- es
- Platforms
- android, windows
- Size
- 40 MB
- License
- Apache-2.0
Vosk 法语小模型
轻量法语离线识别；适合低内存设备
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- fr
- Platforms
- android, windows
- Size
- 42 MB
- License
- Apache-2.0
Vosk 德语小模型
轻量德语离线识别；适合低内存设备
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- de
- Platforms
- android, windows
- Size
- 46 MB
- License
- Apache-2.0
Vosk 俄语小模型
轻量俄语离线识别；适合低内存设备
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- Vosk 0.3.38; Android 10+ or Windows x64
- Languages
- ru
- Platforms
- android, windows
- Size
- 46 MB
- License
- Apache-2.0
SenseVoiceSmall INT8
内置完整推理包；英中日韩粤语识别与自动检测
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Verified
- Runtime
- sherpa-onnx 1.13.4; Android 10+ or Windows x64
- Languages
- auto, zh, en, ja, ko, yue
- Platforms
- android, windows, web
- Size
- 160 MB
- License
- MIT
Omnilingual CTC 300M INT8
内置 1600+ 语言扩展识别；无语义标点时使用规则分句
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Verified
- Runtime
- sherpa-onnx 1.13.4; Android 10+ or Windows x64
- Languages
- auto, 1600+
- Platforms
- android, windows, web
- Size
- 238 MB
- License
- Apache-2.0
Whisper Base 多语
Windows 本地多语识别；使用 sherpa-onnx 原生运行时和共享 Silero VAD
Android · UnavailableWindows · VerifiedWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Windows x64
- Languages
- auto, en, zh, ja, ko, es, fr, de, ru
- Platforms
- windows
- Size
- 208 MB
- License
- MIT
Qwen3-ASR 0.6B INT8
多语高准确率识别；约 879 MB，适合大内存 Android 与 Windows；不内置 APK
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Android 10+ or Windows x64
- Languages
- auto, en, zh, ja, ko, vi, yue
- Platforms
- android, windows
- Size
- 879 MB
- License
- Apache-2.0
punctuation-en
Android · Direct install; auto-verifiedWindows · UnavailableWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Android 10+
- Languages
- en
- Platforms
- android
- Size
- 31 MB
- License
- Apache-2.0
Piper 小雅中文 INT8
Windows 轻量中文本地音色；官方包，需遵守非商业数据许可
Android · UnavailableWindows · VerifiedWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Windows x64
- Languages
- zh-CN
- Platforms
- windows
- Size
- 14 MB
- License
- Non-commercial dataset license; review upstream model card
Piper 超文中文 INT8
Windows 轻量中文本地音色；官方 sherpa-onnx 包
Android · UnavailableWindows · VerifiedWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Windows x64
- Languages
- zh-CN
- Platforms
- windows
- Size
- 14 MB
- License
- CC0 source dataset; derived model redistribution requires review
bergamot-en-zh
Android · UnavailableWindows · VerifiedWeb · Verified
- Runtime
- BergamotTranslatorSharp 0.5.1 on Windows x64 or @browsermt/bergamot-translator 0.4.9 in WebAssembly
- Languages
- en
- Platforms
- windows, web
- Size
- 37 MB
- License
- MPL-2.0
OPUS-MT 英语到中文 INT8
本地 Marian ONNX 翻译；比 ML Kit 更适合英语电影对白，首次下载后完全离线
Android · Direct install; auto-verifiedWindows · VerifiedWeb · Unavailable
- Runtime
- ONNX Runtime 1.27.0; Android 10+ or Windows x64
- Languages
- en
- Platforms
- android, windows
- Size
- 182 MB
- License
- Apache-2.0
bergamot-ja-en
Android · UnavailableWindows · VerifiedWeb · Unavailable
- Runtime
- BergamotTranslatorSharp 0.5.1; Windows x64; requires bergamot-en-zh for Japanese to Chinese
- Languages
- ja
- Platforms
- windows
- Size
- 49 MB
- License
- MPL-2.0
RTT multilingual rule segmentation
Language-aware fallback segmentation for Chinese, Japanese, Korean, and punctuation-poor recognizers.
Android · VerifiedWindows · VerifiedWeb · Verified
- Runtime
- RTT 0.2.0+
- Languages
- auto, zh, yue, en, ja, ko, es, fr, de, it, pt, ru, ar, hi, th, vi, id
- Platforms
- android, windows, web
- Size
- -
- License
- Apache-2.0
Google ML Kit on-device OCR
Android on-device text recognition for selectable screen subtitle regions.
Android · CandidateWindows · UnavailableWeb · Unavailable
- Runtime
- Android 10+; Google ML Kit
- Languages
- zh, en, ja, ko, latin
- Platforms
- android
- Size
- -
- License
- Google ML Kit terms
Google ML Kit translation language packs
Per-language offline translation packs downloaded and managed by ML Kit on Android.
Android · CandidateWindows · UnavailableWeb · Unavailable
- Runtime
- Android 10+; Google ML Kit
- Languages
- multilingual
- Platforms
- android
- Size
- -
- License
- Google ML Kit terms
Tesseract.js OCR language data
Browser OCR candidate for user-selected screen captures; language data stays in browser storage.
Android · UnavailableWindows · UnavailableWeb · Official link only
- Runtime
- Tesseract.js worker
- Languages
- multilingual
- Platforms
- web
- Size
- -
- License
- Apache-2.0
system-voices
RTT 0.2.0
Android · VerifiedWindows · VerifiedWeb · Verified
- Runtime
- RTT 0.2.0
- Languages
- zh, en, ja, ko
- Platforms
- android, windows, web
- Size
- -
- License
- Platform dependent
piper-zh-huayan
sherpa-onnx TTS
Android · Official link onlyWindows · Official link onlyWeb · Official link only
- Runtime
- sherpa-onnx TTS
- Languages
- zh
- Platforms
- android, windows, web
- Size
- -
- License
- Verify upstream model card before redistribution
piper-zh-xiao-ya-int8
sherpa-onnx 1.13.4; Windows x64
Android · UnavailableWindows · VerifiedWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Windows x64
- Languages
- zh-CN
- Platforms
- windows
- Size
- 14 MB
- License
- Non-commercial dataset license; review upstream model card
piper-zh-chaowen-int8
sherpa-onnx 1.13.4; Windows x64
Android · UnavailableWindows · VerifiedWeb · Unavailable
- Runtime
- sherpa-onnx 1.13.4; Windows x64
- Languages
- zh-CN
- Platforms
- windows
- Size
- 14 MB
- License
- CC0 source dataset; derived model redistribution requires review
melotts-chinese
Python local service plugin
Android · UnavailableWindows · CandidateWeb · Unavailable
- Runtime
- Python local service plugin
- Languages
- zh
- Platforms
- windows
- Size
- -
- License
- MIT
cosyvoice3-0.5b
CosyVoice local service plugin
Android · UnavailableWindows · CandidateWeb · Unavailable
- Runtime
- CosyVoice local service plugin
- Languages
- zh, en, ja, ko
- Platforms
- windows
- Size
- -
- License
- Apache-2.0; verify model card and voice rights
Installed models
Downloaded packages use browser OPFS; imported folders remain in their original location.
Download tasks
Tracks download, SHA-256 verification, installation, and real inference validation.
Runtimes
Browser runtimes use OPFS directly. RTT Windows can expose larger local models to Web.
Account center
Review active routes, APIs, models, and account state in one place.
Signed outThis device only
Signed outApp locked · sign in to open the workspace
Data in this browserConnection settings keep a local copy and sync in encrypted form. Model folders and translation content remain on this device.
Configured APIs0
Authorized local models0
Sync statusThis device only
Active capability routesThese are the five capabilities used when a caption or translation task starts.
Developer administrator
Open admin consoleUsers and cloud credentials
Manage RTT users, encrypted credential sync, and audit records. Revealing or changing a key requires recent authentication and an access reason.
Settings
Interface, captions, and browser-local data. Changing the display language updates the entire workspace immediately.
Display language
Caption size
Computer-local models
The Web app directly reads authorized model folders and runs Vosk, sherpa-onnx, Bergamot, and compatible Transformers.js models in browser WASM. The RTT Desktop bridge remains available as an optional desktop reuse route.
Privacy boundaries
Text, files, OCR frames, and capability settings stay in this browser by default. Remote profiles send translation requests to the selected service. Web Speech network behavior depends on the browser implementation.