LiveCaptions-Translator: Real-Time Speech Translation on Top of Windows LiveCaptions
Lightweight and powerful real-time audio/speech translation tool based on Windows LiveCaptions.
At a glance
- What is it?
- LiveCaptions-Translator is a Windows 11 tool that pipes the built-in LiveCaptions text into a translation API, including LLM engines such as Ollama and OpenAI-compatible endpoints. It is a thin layer over Microsoft's recognizer, and that dependency defines both what it does well and where it stops.
- Who is it for?
- LiveCaptions-Translator is for Windows 11 users who already trust the built-in LiveCaptions recognition and want translated subtitles in an overlay or a CSV log, and who have a translation backend ready (Google Translate out of the box, or an LLM endpoint for better handling of partial sentences). It is not for Windows 10, macOS, Android or browser users, because the README lists Windows 11 22H2 or later as a prerequisite and the tool reads LiveCaptions output.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 44 days ago.
- What is it written in?
- Mainly C#, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 30, 2026, and from our analysis. They are not legal advice.
Editorial analysis
What LiveCaptions-Translator solves, and for whom
Windows 11 ships LiveCaptions, an on-device speech recognizer that is cheap to run and, in the project's own words, has "extremely high recognition accuracy." What it does not do is translate. LiveCaptions-Translator takes that recognized text and sends it through a translation API, then renders the result in its own Fluent UI window, an overlay, or a history log. The README summarizes the whole design in one line: "LiveCaptions Translator = Windows LiveCaptions + Translate API."
The audience is narrow but real. You need Windows 11 22H2 or later with LiveCaptions support, and .NET 8.0 or later. That rules out Windows 10, macOS, Android and browser extensions, whatever the search results around this project's name suggest. Within that boundary, the tool targets people who watch or listen to speech in a language they do not read fluently: streams, lectures, meetings, games. The overlay window is explicitly pitched at gaming, videos and live streams, and the README notes it can be made borderless and transparent so it sits on top of whatever is already on screen.
The project also claims you do not need a Copilot+ PC, which matters because Microsoft's own translation features are tied to newer hardware. Here the heavy lifting stays with LiveCaptions and with whichever API you configure.
The mechanism: capture from LiveCaptions, translate through an API
There is no audio pipeline of its own. The README describes the tool as automatically invoking Windows LiveCaptions "without opening separate windows," and after first use LiveCaptions is hidden by default, with a Show/Hide button to bring it back. In other words, the recognizer still runs; the translator wraps it.
Translation is delegated to a backend you choose. The README lists nine engines in a table, split into LLM-based and traditional. LLM-based options are Ollama (self-hosted), OpenAI Compatible API (online) and OpenRouter (online). Traditional engines are Google Translate, DeepL, Youdao, Baidu Translate, MTranServer (self-hosted) and LibreTranslate (self-hosted). The README says two Google Translate engines work out of the box, and it "strongly recommend[s] using LLM-based translation engines, as LLMs excel at handling incomplete sentences and are adept at understanding context." That recommendation is the interesting design claim: live captions arrive in fragments, and a fragment is exactly where a phrase-based engine tends to produce nonsense.
The rest of the data flow is presentation. Recent transcriptions can appear as Log Cards to keep context visible, and original plus translated text is recorded in history, exportable as CSV. The number of sentences in the overlay and the number of Log Cards are both configurable in the settings page.
Installing it and getting a first translation
The README does not document a package manager or a build-from-source path. It points at the Releases page: "Download from Releases and start with a single click." So the install is a binary, and the first real use is configuration rather than code.
Before anything else, confirm the prerequisite. The README's prerequisites table asks for Windows 11 22H2 or later with LiveCaptions support, and .NET 8.0 or later. If you plan to build the solution yourself, the repository root contains LiveCaptionsTranslator.sln and LiveCaptionsTranslator.csproj, so the usual .NET entry point applies:
dotnet build LiveCaptionsTranslator.slnThen open the app, pick a translation engine in the settings, and supply whatever that engine needs (an API key for an online service, or a local endpoint for Ollama or MTranServer). Two warnings in the README matter more than any other step. First, you must change the source language inside Windows LiveCaptions itself, because the translator consumes what LiveCaptions produces. Second, to translate your own voice rather than system audio, enable the "Include microphone audio" option in Windows LiveCaptions settings. With both set, captions appear in the main window and translation follows.
One configuration detail is easy to miss: LiveCaptions is hidden by default after first use. If captions seem to have stopped, check the Show/Hide control before assuming the translator failed.
Where the design breaks down
The hard dependency on Windows LiveCaptions is the main limitation. If LiveCaptions mishears a word, the translator translates the misheard word faithfully. There is no second recognizer to compare against and no confidence signal exposed in the README. The source-language setting inside LiveCaptions is not a convenience; get it wrong and the pipeline produces fluent translations of the wrong text.
Latency is a structural property rather than a bug. Recognition happens first, then an API call, then rendering. A local Ollama model avoids a network round trip but adds inference time; an online LLM adds network time. The README does not publish latency figures, so treat any expectation here as unverified.
Cost and privacy depend entirely on the backend. A cloud LLM or DeepL receives your transcript text. Ollama, MTranServer and LibreTranslate keep it local, but the README does not document what those self-hosted setups require beyond their project links. And the platform boundary is absolute: no Windows 10, no macOS, no Android, no browser extension. If you need translation on a phone or on a Mac, this is the wrong tool, not a tool with a workaround.
LiveCaptions-Translator compared with a dedicated translation app
A standalone speech translation app typically owns the whole chain: it captures audio, runs its own speech model, translates, and displays. That gives it control over the recognizer, which means it can tune for accents, domain vocabulary or low-latency streaming, and it can run anywhere the app is ported.
LiveCaptions-Translator deliberately gives up that control. It inherits Microsoft's recognizer, which means it inherits its accuracy, its language support and its resource profile, and it inherits its availability: Windows 11 only. In exchange, it is small, it does not duplicate a speech model, and it can swap translation backends freely, from a free Google Translate endpoint to a self-hosted Ollama instance. If your problem is that LiveCaptions already transcribes well but you cannot read the output, this is a much smaller intervention than replacing the recognizer. If your problem is that LiveCaptions transcribes badly, this tool cannot help, because it never sees the audio.
Maintenance, licence and upgrade cost
The repository is not archived, and the last push was on 2026-08-17. The most recent release listed is v1.7.1300.1822 from 2026-01-18, with v1.6.1254.2719 before it in 2025-09-27 and v1.6.1253.1718 in 2025-06-17. So the release cadence is irregular, with roughly a nine-month gap between v1.6.1254.2719 and v1.7.1300.1822, while repository activity continued past the last release.
That pattern has a practical consequence: if you install a release binary, you may be running code older than the current master branch. The README does not document an upgrade procedure, a version check, or a rollback path, so treat upgrades as a manual download from Releases. There is also no published compatibility matrix beyond the Windows 11 22H2+ and .NET 8.0+ prerequisites, which means a future LiveCaptions change from Microsoft is a risk this project cannot control.
The licence is Apache-2.0, which is permissive and includes an explicit patent grant. That covers the code in this repository. It does not cover the translation services you connect to: each API (Google Translate, DeepL, Youdao, Baidu, OpenRouter, OpenAI-compatible providers) has its own terms, quotas and billing, and a self-hosted engine carries its own licence. Nothing here is legal advice; check the terms of the specific backend you configure.
Editorial conclusion
LiveCaptions-Translator is for Windows 11 users who already trust the built-in LiveCaptions recognition and want translated subtitles in an overlay or a CSV log, and who have a translation backend ready (Google Translate out of the box, or an LLM endpoint for better handling of partial sentences). It is not for Windows 10, macOS, Android or browser users, because the README lists Windows 11 22H2 or later as a prerequisite and the tool reads LiveCaptions output. Before adopting it, verify three things on your own machine: that LiveCaptions works, that the source language inside LiveCaptions is set correctly, and that your chosen API key or self-hosted endpoint answers. The last one is the boundary that decides everything, because the tool ships no translation model of its own.
Frequently asked questions
How do I use LiveCaptions-Translator?
Download it from the Releases page and start it, then choose a translation engine in the settings and provide whatever that engine needs, such as an API key. You must also change the source language inside Windows LiveCaptions itself, and enable the Include microphone audio option there if you want to translate your own voice.
Does LiveCaptions-Translator run on Windows 10?
No. The README lists Windows 11 22H2 or later with LiveCaptions support as a prerequisite, because the tool works from Windows LiveCaptions output. It also requires .NET 8.0 or later.
Which translation engines does LiveCaptions-Translator support?
The README lists Ollama, OpenAI Compatible API and OpenRouter as LLM-based engines, and Google Translate, DeepL, Youdao, Baidu Translate, MTranServer and LibreTranslate as traditional engines. Two Google Translate engines work out of the box, and the README recommends the LLM-based ones for handling incomplete sentences.
Does LiveCaptions-Translator work without an internet connection?
Only if you use a self-hosted engine, since the online engines send text to a remote API. The self-hosted options named in the README are Ollama, MTranServer and LibreTranslate.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/sakirinn-livecaptions-translator)