Who LiveCaptions-Translator is for#
Gamers and streamers following foreign-language content
Set up the transparent overlay window to display translated subtitles directly over a game or stream without a separate window. The click-through mode ensures the overlay does not interfere with mouse input. Configure once per session; the overlay persists across window focus changes.
Skip if:
You are watching content with existing subtitle tracks (SRT files, native platform subtitles). LiveCaptions-Translator targets audio sources that lack pre-existing subtitles or caption tracks.
Professionals attending multilingual video calls
Enable 'Include microphone audio' in Windows Live Captions to transcribe and translate incoming speech during meetings. The history tab records the full transcript with original and translated text side by side, and exports to CSV for post-meeting reference.
Skip if:
Your meeting platform (Zoom, Teams, Google Meet) already includes real-time translation in your target language. Use the platform's native feature to avoid audio routing complexity.
Students watching foreign-language lectures or videos
Run the tool alongside any video player or browser tab. Log Cards display multiple recent sentences so you can maintain context even when the subtitle overlay only shows the latest segment. Export history as CSV at the end of a session for later review.
Skip if:
The video platform provides captions or auto-generated subtitles in your target language. YouTube auto-translate covers the most common language pairs at no cost.
Developers testing self-hosted LLM translation quality
LiveCaptions-Translator supports Ollama as a translation backend, making it a practical test environment for evaluating local LLM translation on real conversational input. Configure any Ollama model and compare output quality against cloud APIs in a real-use scenario rather than a static benchmark.
Skip if:
You need a production translation pipeline for batch processing or server-side use. This is a desktop application for individual Windows users, not a translation microservice.
The problem it solves#
Real-time speech translation has been a pay-per-use service for most Windows users. Platforms like Wordly and KUDO charge per-minute or per-event for conference interpretation. DeepL Voice requires a paid subscription. Microsoft's own translation layer for Windows Live Captions is locked to Copilot+ PCs, leaving anyone on standard Windows 11 hardware without a built-in path.
The deeper pain is fragmentation: a gamer might use one app for subtitles, a professional another for meeting transcription, and a student yet another for video playback. None of these integrate with the OS-level caption layer, and most route audio to remote servers with no local option. For users who need persistent, context-aware translation across everything they hear on their PC, no commercial tool offers both broad engine support and self-hosted privacy in a single, lightweight package.
How it solves it#
Nine Translation Backends Including Self-Hosted LLMs
Supports Ollama (self-hosted LLM), OpenAI-compatible APIs, OpenRouter, Google Translate, DeepL, Youdao, Baidu Translate, MTranServer (self-hosted), and LibreTranslate (self-hosted). LLM-based backends are recommended because they maintain conversational context across mid-speech fragments that traditional machine translation engines mishandle.
Transparent Click-Through Overlay Window
Displays translated subtitles as a borderless, transparent overlay that can be made fully click-through, so it does not interfere with mouse input during gaming or video watching. Background color, subtitle color, font size, and transparency are all configurable. The number of sentences shown simultaneously is adjustable in settings.
Translation History with CSV Export
Records both the original transcribed text and the translated output for every session. Export the full history as a CSV file, useful for meetings, lectures, and discussions where a searchable reference matters afterward.
Silent Windows Live Captions Integration
Automatically invokes Windows Live Captions in the background without opening a separate window. Reads Live Captions output directly, inheriting Windows 11's on-device speech recognition accuracy without additional audio processing overhead. Microphone translation is enabled by toggling 'Include microphone audio' in Windows Live Captions settings.
Log Cards for Conversational Context
Displays recent transcription segments as Log Cards in a sidebar, showing multiple previous sentences so you can follow the flow of a conversation even when the subtitle overlay only shows the current segment. The number of cards shown is configurable in settings.
Strengths and trade-offs#
Strengths
- Works on Any Windows 11 PC, Not Just Copilot+Microsoft's native Windows Live Captions translation requires a Copilot+ PC with an NPU. LiveCaptions-Translator bypasses that restriction by handling translation through a separate configurable API, working on any Windows 11 22H2+ machine with no hardware upgrade needed.
- LLM-Backed Translation for Conversational AudioTraditional machine translation engines struggle with mid-sentence fragments typical in live speech. LLM-based backends (Ollama, OpenRouter, OpenAI-compatible) maintain context across segments and produce more natural output for conversational audio. The project explicitly recommends these over traditional MT backends for best results.
- Fully Private with Self-Hosted BackendsThree backends (Ollama, MTranServer, LibreTranslate) run entirely on local infrastructure. Windows Live Captions handles speech recognition on-device. For users with privacy requirements, no audio or text needs to leave the machine when using these backends.
- Strong Adoption for a Young ProjectReached 3,518 GitHub stars since its September 2024 launch, with 248 forks and active maintenance through July 2026. The Trendshift recognition and consistent commit history indicate genuine community adoption for a project under two years old.
Trade-offs
- -Windows 11 Only, No Linux or macOS SupportThe tool depends on Windows Live Captions, a feature exclusive to Windows 11 version 22H2 and later. There is no Linux, macOS, or Windows 10 path. Users on other operating systems have no supported alternative using this tool.
- -Multi-Step Initial Configuration RequiredBefore first use, you must open Windows Live Captions manually, consent to on-device voice processing, download language files, and set the position to 'Overlaid on screen.' The README marks this last step as critical: skipping it causes a display bug. The steps are spread across Windows Settings menus and take several minutes on a fresh system.
- -Translation Quality Varies by Backend and Language PairSpeech recognition accuracy depends on the Windows Live Captions language model configured separately in Windows Settings. Translation quality depends on the chosen backend. Cloud LLM APIs incur costs; the free Google Translate endpoint may produce lower quality for less common language pairs. Self-hosted Ollama models require a capable GPU for good real-time performance.
LiveCaptions-Translator vs alternatives#
LiveCaptions-Translator vs Microsoft Translator (Windows Live Captions)
Microsoft ships a translation layer inside Windows Live Captions on Copilot+ PCs with an NPU. On standard Windows 11 hardware, that feature is unavailable.
| Feature | LiveCaptions-Translator | Microsoft Translator (Windows) |
|---|---|---|
| License | Apache-2.0 | Proprietary |
| Hardware requirement | Any Windows 11 22H2+ | Copilot+ PC (NPU required) |
| Translation backends | 9 (Ollama, DeepL, Google, and more) | Microsoft Translator only |
| Self-hosted translation | Yes (Ollama, MTranServer, LibreTranslate) | No |
| Overlay window | Yes (transparent, click-through) | No |
| History export | Yes (CSV) | No |
LiveCaptions-Translator is the better choice for users on standard Windows 11 hardware and for those who want backend flexibility or self-hosted privacy. Microsoft's native translation is worth considering only if you already own a Copilot+ PC and want zero-configuration translation with no third-party app to install.
LiveCaptions-Translator vs DeepL Voice
DeepL Voice is a paid commercial service for real-time meeting translation, aimed primarily at enterprise conference use. It routes audio to DeepL's cloud infrastructure.
| Feature | LiveCaptions-Translator | DeepL Voice |
|---|---|---|
| License | Apache-2.0 | Proprietary |
| Pricing | Free (API costs may apply) | Paid subscription |
| Self-hosting | Yes (via compatible backends) | No |
| Platform | Windows 11 only | Multi-platform |
| History export | Yes (CSV) | Limited |
LiveCaptions-Translator is the better choice for individual Windows users who want persistent subtitle translation across all audio sources at no per-seat cost. DeepL Voice makes more sense in enterprise contexts that require multi-platform support and managed infrastructure. DeepL's language quality for European language pairs is strong; if that is the priority and budget is not a constraint, DeepL Voice is worth evaluating.
LiveCaptions-Translator vs Google Gemini Live Translate
Google Gemini Live Translate is a real-time translation feature tied to Google's AI subscription, primarily targeting mobile and Google Meet scenarios. It is not a standalone Windows desktop application.
| Feature | LiveCaptions-Translator | Google Gemini Live Translate |
|---|---|---|
| License | Apache-2.0 | Proprietary |
| Desktop (Windows) support | Yes | Limited / browser-based |
| Offline or self-hosted | Yes (with Ollama) | No |
| Backend choice | 9 options | Google only |
| Persistent overlay | Yes (transparent, click-through) | No |
LiveCaptions-Translator wins for desktop Windows workflows where you want a persistent overlay across all audio sources. Google Gemini Live Translate is stronger if you are in Google's ecosystem and need mobile or Google Meet integration. LiveCaptions-Translator also supports Google Translate as one of its nine backends, so you are not giving up Google's translation quality by choosing this tool.
What it's built on#
- Languages
- C#
FAQ#
Does LiveCaptions-Translator work on Windows 10?
No. The tool depends on Windows Live Captions, which is only available on Windows 11 version 22H2 and later. Windows 10 does not include Live Captions, so there is no supported path on that OS.
Which translation backend should I use?
The README recommends LLM-based backends (Ollama, OpenRouter, or an OpenAI-compatible API) because they handle mid-sentence fragments and conversational context better than traditional machine translation engines. For zero-cost use with no additional setup, the built-in Google Translate endpoint works out of the box. If you want fully private translation with no data leaving your machine, configure Ollama with a locally hosted model.
Can I use LiveCaptions-Translator without an internet connection?
Yes, for the translation layer, if you configure Ollama or MTranServer as local backends. Windows Live Captions handles speech recognition on-device. An internet connection is only required when using cloud-based translation APIs like Google Translate, DeepL, or OpenRouter.
Is a Copilot+ PC required to run LiveCaptions-Translator?
No. Microsoft's own Windows Live Captions translation feature requires a Copilot+ PC with an NPU. LiveCaptions-Translator does not: it works on any Windows 11 22H2+ machine by handling translation through a separate configurable API backend, bypassing the hardware restriction entirely.
How does translation history work?
The tool records both the original transcribed text and its translation for every session. You can view this in the history tab during use and export the entire record as a CSV file, useful for meetings or lectures you want to review afterward.
Similar open-source tools#
whishper
Local speech-to-text transcription and subtitling with a web UI
CLI-Anything
Empower AI agents with agent-native CLIs
discli
CLI for building AI agents and automating Discord servers
Zed
High-performance open source code editor with built-in AI
dograh
Open source voice agent builder with telephony and BYOK AI support
openwhispr
Local-first voice dictation with Whisper, offline mode, and AI cleanup