Speech recognition runs on your machine
Transcription happens locally on your GPU. Polyvox does not send your audio anywhere and does not write recordings to disk.
A Windows desktop application that listens to your microphone and to the audio your PC plays, transcribes the speech on your own GPU, translates it, and puts the result where you can read it: the VRChat chatbox, a desktop overlay, and a panel inside SteamVR.
Transcription happens locally on your GPU. Polyvox does not send your audio anywhere and does not write recordings to disk.
Translation is the step that can leave your machine, and which service handles it is your choice. The privacy page lists every path that carries data off your PC.
The VRChat chatbox over OSC, a desktop overlay you can move and resize, and a panel rendered inside SteamVR.
Furigana, pinyin and pitch accent are rendered inline, so a line you cannot read yet is still a line you can sound out.
Polyvox itself is free to download and use, including local transcription, translation through a service you supply your own key for, the chatbox output, and the overlays. The subscription adds four things.
| Included | What it does |
|---|---|
| Study | Polyvox builds a deck out of the words you actually met while talking to people, ranks them by how often they came up, and schedules them for review. Reading aids for Japanese and Chinese are built in. |
| Lesson plans | From that deck, Polyvox generates lessons and speaking practice aimed at the words you keep meeting and keep missing, rather than at a fixed course order. |
| Polyvox translation | Translation through a service we operate, so you do not have to register for a translation provider or manage an API key. Includes up to 20,000 translated lines per month. |
| AI assistant | Ask about a line you just heard: what a word means, why the grammar works that way, how to answer. Metered against the same monthly allowance. |
Every other engine stays available and unchanged. If you already pay a translation provider directly, you can keep using it on the free tier.
Everything in the paid tier, on a month to month basis. Stop whenever you like.
The same thing, paid once a year. Two months cheaper than paying monthly.
All prices are in United States dollars (USD) and are shown before any tax that applies where you live. There is no free trial that turns into a charge, and no setup fee.
A subscription bought with a card renews automatically at the price above until you cancel it. A subscription bought with WeChat Pay or Alipay does not renew automatically: those wallets do not support merchant-initiated recurring charges, so you buy a fixed number of months up front and start a new purchase when you want to continue. Nothing is ever charged to a wallet without you approving that specific payment.
You can cancel at any time, with no cancellation fee and no minimum term. Cancel by emailing terry61295@gmail.com from the address you bought with, or from the billing link in your purchase receipt. Cancellation stops the next charge. The features you paid for keep working until the end of the period you already paid for, and are not cut off early.
When a subscription ends, Polyvox does not lock. The application keeps running and keeps transcribing and translating with an engine you supply. Only the four items listed above stop being available. See the refund policy for how to get money back.
| Requirement | Detail |
|---|---|
| Operating system | Windows 10 or Windows 11, 64-bit |
| Graphics | A GPU with Vulkan support. Transcription is GPU work, and speed depends on the card. |
| VR | Optional. The desktop overlay and the chatbox work without a headset. |
No. Audio is processed in memory and discarded. Polyvox does not write your microphone or system audio to disk. Two narrow exceptions exist, both switched off unless you turn them on and both explained on the privacy page.
Polyvox can transcribe the audio your PC is playing, which includes other people speaking in VRChat. Please be aware of what the people around you expect, and of the recording and consent laws where you live. Polyvox is a tool for understanding a conversation you are part of, not for surveillance.
Transcription and translation cover a wide set of languages. The reading aids (furigana, pinyin, pitch accent) are built for Japanese and Chinese.
Email terry61295@gmail.com. Please include your Polyvox version, found under Settings. We answer within two business days.