Captions alone do not join the conversation
In VC, being able to read is not the same as being able to answer with your voice.

VoxShift transcribes and translates your microphone input, then outputs translated speech with reference-voice synthesis for Discord, game voice chat, OBS, and meeting apps.
Partner subtitles translate the other person's speech into text, so conversations with friends, viewers, and teammates are supported by both voice and captions.

Supported Languages
Caption translation helps you read, but it does not let others hear you speak. Generic TTS removes your identity, and voice changers do not translate language. VoxShift is built for the gap between translation tools and real voice chat.
In VC, being able to read is not the same as being able to answer with your voice.
Even when the words are translated, a generic synthetic voice can feel detached from you.
VoxShift is designed around virtual mic output, monitoring, OBS captions, and game-friendly routing.
Speaks the translation in a voice close to yours, based on a short optional recording or your live voice.
Fast starts translating before you finish the sentence; Quality waits for more context. Choose on the Live tab.
The Live tab shows what is being said in Now and finished lines in History, newest first.
Sends translated voice to Discord, games, and OBS through a virtual audio device. Monitor lets you hear it yourself.
Your speech is output as voice, while the other person's speech (call, game, or video audio) is translated as subtitles.
Shows subtitles in the built-in subtitle window or sends original and translated captions to an OBS text source.
Walks through display language, quality for your PC, languages, microphone and output, and an optional voice recording.
Checks your PC's GPU memory and CPU, then picks Standard, Low VRAM, or Ultra-low VRAM. Change it in Performance → Manual settings.
The Live tab checks whether Live can start with your current settings; the Performance tab shows estimated VRAM.
The bidirectional mode is asymmetric to reduce GPU load and latency. Your speech is voiced, while the other person's speech is translated into subtitles.
Voice output
Subtitle translation
These screenshots are captured from the current VoxShift app running on a local PC. Instead of a browser demo that hides PC requirements, the free version (Steam Demo / Microsoft Store free trial) lets you check your GPU, audio devices, and whether Live can start on your own PC.



The core is still reference-voice translation. Around it, Soundboard, voice polish, and routing views make it easier to imagine how VoxShift fits Discord, game VC, and OBS after purchase.
Core
Reference voice translation + subtitles
Support
Soundboard / EQ / Routing

Soundboard cues
ReadyVoice polish
One-shot cues
Keep short confirmations and reactions ready, so waiting for a translation does not stop the flow of VC.
Noise / EQ / Clipping protection
Translated voice is easier to use when cleanup, EQ, and clipping protection are part of the same workflow.
VC / Monitor / OBS
Check virtual mic, monitor output, and OBS captions as one route to reduce setup mistakes.
Talk with overseas friends in a foreign language using your own voice.
Send callouts, tactics, and reactions as translated speech.
Add translated voice and captions for international viewers.
Usable in meetings, though the main focus is VC, streaming, and games.
These pages break down the main ways people look for voice translation: Discord, game voice chat, OBS captions, reference voice, Windows setup, and local-first processing.
Discord Voice Translator
Translate Discord voice chat on Windows with reference-voice speech output, virtual mic routing, partner subtitles, and a free version to test first.
Game Voice Chat Translation
Translate your microphone for game voice chat, Discord calls, OBS streams, and international teams with VoxShift for Windows. Free version on Steam and Microsoft Store.
OBS Translation Captions
Add translated captions and reference-voice translated speech to OBS streams with VoxShift for Windows. Supports OBS WebSocket text source updates.
Reference Voice Translation
Translate speech while keeping a voice close to your reference voice. VoxShift outputs reference-voice translated speech for Discord, game VC, and OBS on Windows.
Real-Time Voice Translation
Near-real-time AI voice translation for Windows. Translate your microphone into reference-voice speech for Discord, game VC, OBS, and meetings.
Live Translation Captions
Translate microphone or desktop audio into real-time AI subtitles for a built-in subtitle window and OBS captions. Test latency with the VoxShift free version.
Windows Voice Translator
VoxShift is a Windows voice translator that combines microphone input, speech recognition, translation, reference-voice synthesis, OBS captions, and virtual audio routing.
System Audio Translator
Translate Windows system and desktop audio into live captions with VoxShift. Test YouTube, calls, streams, audio routing, source limits, and latency with the free version.
Meeting Voice Translator
Use VoxShift with Zoom, Teams, and Google Meet through your Windows audio setup. Test translated speech, captions, latency, and routing with the free version.
Local / Privacy-first Voice Translation
Privacy-focused Windows voice translation. By default, VoxShift does not send audio, reference voice, captions, or translated text to operator servers.
Virtual Microphone Translator
Route translated speech to Discord, game voice chat, OBS, and call apps through Windows audio setup. Test routing with the VoxShift free version.
Voice Translator for Streamers
Use translated speech and OBS translated captions for Twitch, YouTube Live, Kick, and international streams. VoxShift is Windows voice translation software for streamers.
Microsoft Store Voice Translator
VoxShift is a Windows voice translator on Microsoft Store. Use the free trial to check GPU, VRAM, audio devices, Discord routing, and OBS captions before buying.
Soundboard Voice Translator
VoxShift focuses on reference-voice translation, with Soundboard cues, voice polish, OBS captions, and routing checks for Discord, game VC, and streams.
VoxShift handles AI models, virtual audio, microphones, and output routes. The setup wizard walks through the checks in order, so configuration issues are easier to find before Live starts.

Free
Download the free versionOne-time purchase
Buy nowThe full version is sold on Steam (Early Access) and the Microsoft Store, with no subscription. Each store is a separate product and license: buying on one does not unlock the other. See each store page for the price.
Live voice translation requires an NVIDIA CUDA GPU (Turing or newer) with 4 GB VRAM or more. Please try the free version on your PC before buying.
Live translation with reference-voice synthesis requires an NVIDIA CUDA-capable GPU, Turing or newer (GeForce GTX 16 / RTX 20 series or later), with 4 GB VRAM or more. On Radeon, Intel GPU, or no-GPU systems, Live voice generation is unavailable. VoxShift checks your PC and picks a GPU memory mode automatically; to choose it yourself, use Performance → Manual settings. Optional higher-quality models need more VRAM.
| Mode | VRAM | Use Cases |
|---|---|---|
| Setup / checks | Not required | Setup wizard, audio checks, and diagnostics, also on Radeon / Intel / no-GPU PCs |
| Standard | From about 5.8 GiB | PCs with enough GPU memory |
| Low VRAM | About 3.8 GiB | PCs with less GPU memory |
| Ultra-low VRAM | About 2.4 GiB | PCs with very little GPU memory; translation runs on the CPU and can be slower |
No. VoxShift is semi-real-time voice translation, and delay varies with your GPU, VRAM, sentence length, and settings. Fast mode starts translating before you finish the sentence.
Live voice translation requires an NVIDIA CUDA GPU, Turing or newer (GeForce GTX 16 / RTX 20 series or later), with 4 GB VRAM or more. When GPU memory is tight, VoxShift picks Low VRAM or Ultra-low VRAM mode automatically. On Radeon, Intel GPU, or no-GPU PCs, only setup and diagnostics are available.
The Steam Demo and the Microsoft Store free trial have the same limits: each Live session stops after 10 minutes (you can start again), translation output is English only, the light model set is fixed, and the soundboard holds up to 3 items. Setup, voice recording, device checks, and the OBS link can all be tried.
Yes. Partner subtitles listen to the other person (call, game, or video audio) and show their speech as translated subtitles. It is not read aloud.
No. Overlapping speech may reduce accuracy. Turning on “Only voice my own speech” keeps other voices your mic picks up from being spoken.
No. It aims for a voice close to yours, not a perfect copy. A short recording of you reading one sentence aloud helps it sound more like you.
No. Voices without consent are prohibited.
Live voice translation requires an NVIDIA CUDA GPU (Turing or newer) with 4 GB VRAM or more. Please try the free version on your PC before buying.