Home › Real-time interview translator

Real-time interview translator — see what they said, instantly

If English is not your first language, the hardest moment of a job interview is not the technical question. It is the one second after the question — the silent gap where your brain is still parsing accent, vocabulary and idiom while the interviewer is already waiting for an answer. A real-time interview translator closes that gap. It listens to the call, transcribes every sentence as it is spoken, and shows you the translation in your native language as live subtitles on your screen, usually within two seconds.

This article explains what a real-time interview translator is, who actually needs one, how the live audio → ASR → translation → overlay pipeline works, which conferencing platforms are supported, and how a privacy-first implementation differs from a generic AI copilot. Everything here describes the Live Interview Translator that ships with the Quest2Offer desktop app.

Who actually needs a real-time interview translator

The honest answer: anyone who has ever sat in a video call and thought "I understand every word individually, but I lost the whole sentence." Concretely, three groups benefit most:

• Non-native English speakers interviewing at international companies. You have a strong portfolio, you can write English documentation all day, but a fast-talking American product manager with a Boston accent at 11pm your time is a different challenge. Live subtitles in your native language let you focus on the meaning of the question instead of decoding pronunciation.
• ESL job seekers relocating to a new market. Engineers, designers and product people moving from Latin America to the US, from Eastern Europe to Germany, from Central Asia to the UK. A real-time interview translator turns the conversation into a written, translated transcript you can re-read at your own pace.
• Anyone interviewing in a second or third language. A French candidate interviewing in German, a Spanish candidate interviewing in English, a Russian-speaking candidate interviewing in French. Cross-language interviews are increasingly common, and a generic Zoom caption track does not translate — it only transcribes.

The pattern across all three groups is the same: comprehension is not the problem in isolation, but comprehension under interview-grade time pressure is. Live, translated subtitles remove the time pressure on the listening half of the conversation so you can put your full energy into the answering half.

How it works: mic + system audio → ASR → translation → overlay

The pipeline behind a real-time interview translator is conceptually simple but every link in the chain has to be fast. Here is what happens between the interviewer's voice and the translated subtitle on your screen, step by step.

1. Audio capture. The desktop app captures two streams in parallel: your microphone (for you, optional) and the system audio (everything the operating system is playing — which includes the interviewer's voice through Zoom or Meet). On macOS this uses ScreenCaptureKit; on Windows it uses WASAPI loopback. Both produce a clean 16 kHz mono PCM stream without going through the Zoom plugin API.
2. Streaming to ASR. The PCM stream is chunked into short segments, base64-encoded and pushed over a websocket to a self-hosted ASR server running faster-whisper. Transcripts come back incrementally — partial first, then finalized — so the interface can update progressively as the sentence unfolds.
3. Translation. Each finalized chunk is sent to a self-hosted Qwen3.5 LLM with a bilingual prompt that asks for the target language and the input language in a single response. Doing both in one LLM call removes a network round-trip and keeps the user-perceived latency under two seconds.
4. Overlay rendering. The translated text is pushed back through the websocket and rendered in a small, always-on-top desktop window that you can position anywhere on your screen — typically just below the video tile so your eyes barely move.

The whole loop runs continuously for the length of the call. Nothing is recorded to disk, nothing is sent to OpenAI or Anthropic or Google. The model that does the heavy lifting is the same one we use for the rest of Quest2Offer's AI features, so the translation also benefits from your resume context and the vacancy description — technical terms come out right instead of being translated literally.

Platforms: Zoom, Google Meet, Teams and beyond

The single biggest practical question about a real-time interview translator is "does it work with my interview platform?" Because Quest2Offer captures audio at the OS level, the answer is yes for every common video-conferencing tool, and it does not require any plugin, browser extension or admin permission inside the conferencing app.

• Zoom — works on both the native desktop client and Zoom-in-browser. The translator hears the call through system audio loopback, so there is no Zoom marketplace approval needed.
• Google Meet — works in any modern browser. Meet's own captions are English-only and miss most non-American accents; the translator handles accented English and produces native-language output.
• Microsoft Teams — works on both the desktop and web client. Teams' built-in translation is gated behind enterprise licensing; the desktop translator is not.
• Other tools — Webex, Around, Whereby, Discord, Slack huddles, in-browser hiring platforms like HireVue and CodeSignal Interview, even phone calls routed through a softphone. Anything the OS can play, the translator can hear.

Live captions inside Zoom and Meet have improved over the past two years, but they share the same three limits: English-only, no translation, and accuracy collapses on accented speech. A dedicated real-time interview translator solves all three because it owns the entire pipeline end to end.

For a deeper walkthrough of the Zoom and Meet setup specifically, see our live subtitles for Zoom and Google Meet guide.

Privacy: why "self-hosted" matters for interview audio

An interview is one of the most sensitive conversations you will ever have on your computer. Any product that processes that audio has to take privacy seriously, and "we don't store it" is not enough — the question is where the audio goes during the call.

Quest2Offer's Live Interview Translator streams audio to our own GPUs. There is no third-party API in the audio path — not OpenAI Whisper, not Google Cloud Speech, not Anthropic. The ASR model runs on hardware we control, the translation LLM runs on hardware we control, and the connection between them never leaves the cluster. After the call, no audio is persisted: the websocket closes, the in-memory buffers are discarded, and the only thing that remains is the count of seconds used against your plan quota.

For the same reason, the translator does not need any permissions inside Zoom, Meet or Teams. It is a separate desktop application that captures audio from your machine. The interviewer sees you on camera and nothing else. There is no notification, no banner, no "this call is being transcribed" message — because none of that is happening on their side of the call.

How it fits with the rest of Quest2Offer

The live translator is the on-the-call piece, but a job interview is decided in the weeks before the call as much as in the call itself. Quest2Offer pairs the translator with two other tools that share the same context — your resume, your target role, the specific vacancy:

• The AI career consultant reviews your resume, builds a personalized study plan and answers strategy questions ("which of these three offers should I take?", "is this salary band realistic?") in your native language.
• The AI mock interview lets you rehearse the actual conversation in your native language with scored feedback on each answer, so the real call is the second time you have answered each question — not the first.

Because the same context flows into all three tools, the live translation on the day of the interview already "knows" the technical vocabulary of your target role. Generic translators turn "shard the read replica" into something between hilarious and unusable; a context-aware translator keeps it as the sentence the interviewer actually said.

Frequently asked questions

What is a real-time interview translator?

A desktop application that listens to the audio of your video call, transcribes each spoken sentence with automatic speech recognition, translates it into your native language, and shows the result as live subtitles on your screen — all within about two seconds of the words being spoken.

Is it legal to use a translator during a job interview?

Yes. Using assistive subtitle software on your own machine is no different from having a personal interpreter at your shoulder. Nothing is sent to the interviewer, and Quest2Offer does not record the call or store the audio.

Does it work with Zoom, Google Meet and Microsoft Teams?

Yes. Because the translator captures system audio at the operating-system level, it works with any video-conferencing app: Zoom, Google Meet, Microsoft Teams, Webex, Around, Whereby and even browser-based interview platforms.

Will the interviewer notice that I am using a translator?

No. The translator is a separate desktop window on your computer. It is not a plugin for Zoom or Meet and it does not appear in the shared screen unless you explicitly share that window. The interviewer only sees you on camera.

How much latency is there between the question and the translation?

Typically under two seconds end-to-end: audio is streamed to our GPUs, transcribed by faster-whisper, then translated by a self-hosted Qwen3.5 model, and the result is pushed back over the websocket and rendered on screen.

macOS and Windows · works in Zoom, Meet, Teams · no recordings, self-hosted LLM