ChatGPT talks. Gemini talks. Almost none of them caption themselves. Here's how to read any AI chatbot's voice reply in real time — translated, if you need it.
AI chatbots talk now. ChatGPT has a voice mode, Gemini has Live, and a growing list of assistants answer out loud instead of typing. None of them reliably caption their own reply for you — the transcript you see, if there is one, is often just your side of the conversation. If you need to read what an AI chatbot said instead of just hearing it, something listening to your speakers has to produce that text. That's true whether you're deaf or hard of hearing, working in a language that isn't your first, or simply in a room where audio isn't an option.
Voice interfaces spread fast this year. ChatGPT's Advanced Voice Mode, Google's Gemini Live, and a wave of assistant apps now answer out loud by default, the same way a phone call or a video meeting has audio. But a phone call and a video meeting both eventually grew captions, because a platform decided to own the transcript. An AI chatbot's voice reply usually has no equivalent: the app is generating speech, not committing to show you the words behind it.
Some chatbot apps show a running transcript of your own side, because they have to turn your speech into text to answer you at all. Fewer show a clean transcript of their own reply, and whether they do varies by app, by platform, and by whether you're on a phone, a browser, or a desktop client — and that can change with any update. If you need the reply as text every time, the safer approach is the one that already works for an uncaptioned call or meeting: something listening to your speakers turns the audio into a transcript, independent of whatever the chatbot app itself shows.
A deaf or hard-of-hearing user has no way to follow a voice-only reply unless something puts it in text — the same gap that makes an uncaptioned meeting a barrier, not a convenience.
Chatbot voice modes are popular for practicing a second language, but hearing a fast reply in an unfamiliar accent is harder than reading it — and reading it translated is easier still.
Creators who talk to an AI chatbot on stream need captions for viewers watching muted, the same reason they'd caption any other stream audio.
Playing a chatbot's voice reply out loud isn't always an option. Captions let you keep talking to it with the sound off.
Unicaption doesn't integrate with ChatGPT, Gemini, or any specific chatbot — it doesn't need to. It reads system or tab audio directly, the same way it captions a phone call or a streaming platform, which means it captions whatever your computer is playing, including a chatbot's spoken reply, without joining anything or requiring the chatbot app to cooperate.
Start a voice conversation in ChatGPT, Gemini, or whichever assistant you use — in its app, or in a browser tab.
In Unicaption, choose system or tab audio so it reads the chatbot's voice directly — not your microphone, which would also try to caption your own side of the conversation.
Choose the language you expect the reply in, or leave auto-detect on. Set a target language if you want a live translation next to the original.
Ask your question, let the chatbot answer out loud, and read the caption as it appears — moments behind the reply. Add any unusual terms to the custom dictionary first if the topic is technical.
AI chatbot voice modes typically reply in whatever language you spoke in — useful if you're practicing that language, less useful if you asked a quick question in your second language and want the nuance of the answer in your first. Live translation solves that without changing how you talk to the chatbot at all: you speak to it exactly as you normally would, and the caption on your side shows the reply translated.
It also works in reverse for practice: read the original alongside the translation, catch words you didn't recognize by ear, and build vocabulary from an answer you already understand the meaning of.
| Your situation | Best setup | Why |
|---|---|---|
| Chatbot voice mode running in a browser tab | Tab audio | Cleanest source — captions only the chatbot's side |
| Chatbot app on desktop, not a browser tab | System audio | Reads everything your computer plays, including a native app |
| Chatbot on your phone, played through a laptop | Microphone, close range | Keep the phone speaker near the mic and the room quiet |
| Practicing a language or reading in translation | System or tab audio + live translation | Original and translated reply side by side |
Yes. ChatGPT doesn't need to cooperate for this to work — run Unicaption alongside it with system or tab audio as the source, and it captions whatever ChatGPT says out loud, independent of whatever transcript ChatGPT's own interface does or doesn't show.
Yes, and with any voice chatbot, current or future. Unicaption reads audio your computer is playing rather than integrating with a specific app, so it works the same way for Gemini Live, a phone assistant played through a laptop mic, or any voice interface that comes next.
No — Unicaption doesn't join, connect to, or interact with the chatbot app in any way. It listens to audio your own computer is already playing, the same as it would for a video call or a stream, and that audio is never stored.
Yes. Unicaption shows the original transcription and a live translation side by side, in 60+ languages, so you can talk to a chatbot in one language and read its reply in another, moments behind the spoken answer.
Live captions and translation for any chatbot's spoken reply your computer can hear. 30 free minutes every week — no credit card.
Start Free Trial →